diff --git a/litellm/llms/anthropic/common_utils.py b/litellm/llms/anthropic/common_utils.py index 3297aa95715..9461e40cf2e 100644 --- a/litellm/llms/anthropic/common_utils.py +++ b/litellm/llms/anthropic/common_utils.py @@ -440,6 +440,16 @@ class AnthropicModelInfo(BaseLLMModelInfo): """ return AnthropicModelInfo._supports_model_capability(model, "thinking_always_on", custom_llm_provider) + @staticmethod + def _supports_legacy_thinking(model: str, custom_llm_provider: str) -> bool: + """Whether ``model`` is an adaptive-thinking model that still accepts legacy + ``thinking.type=enabled`` with ``budget_tokens`` (the Claude 4.6 family). + The model cost map is authoritative: an explicit ``supports_legacy_thinking`` + entry resolved under ``custom_llm_provider``, or a ``fallback_generalizations`` + rule for unmapped 4.6 ids. Absent flag means the model rejects the legacy shape. + """ + return AnthropicModelInfo._supports_model_capability(model, "supports_legacy_thinking", custom_llm_provider) + @staticmethod def maybe_drop_disabled_thinking( model: str, diff --git a/litellm/llms/anthropic/experimental_pass_through/messages/transformation.py b/litellm/llms/anthropic/experimental_pass_through/messages/transformation.py index adabfa2d62d..032bf0130ce 100644 --- a/litellm/llms/anthropic/experimental_pass_through/messages/transformation.py +++ b/litellm/llms/anthropic/experimental_pass_through/messages/transformation.py @@ -379,13 +379,19 @@ class AnthropicMessagesConfig(BaseAnthropicMessagesConfig): def _translate_legacy_thinking_for_adaptive_model( model: str, optional_params: dict, custom_llm_provider: str ) -> None: - """Translate legacy ``thinking.type=enabled`` to adaptive for 4.6/4.7. - Caller-provided ``output_config.effort`` is never overridden. + """Translate legacy ``thinking.type=enabled`` to adaptive for the + adaptive-thinking models that reject it (4.7+ and the 5 families). + Models flagged ``supports_legacy_thinking`` (the 4.6 family) accept the + legacy shape natively, so it is forwarded verbatim and the caller's + ``budget_tokens`` cap keeps applying. Caller-provided + ``output_config.effort`` is never overridden. """ from litellm.llms.anthropic.chat.transformation import AnthropicConfig if not AnthropicModelInfo._is_adaptive_thinking_model(model, custom_llm_provider): return + if AnthropicModelInfo._supports_legacy_thinking(model, custom_llm_provider): + return thinking: Final = optional_params.get("thinking") if not isinstance(thinking, dict) or thinking.get("type") != "enabled": return diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index 79d9e579de7..f0ed171ad96 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -385,9 +385,9 @@ }, "amazon.nova-2-pro-preview-20251202-v1:0": { "cache_read_input_token_cost": 5.46875e-07, - "input_cost_per_token": 2.1875e-06, - "input_cost_per_image_token": 2.1875e-06, "input_cost_per_audio_token": 2.1875e-06, + "input_cost_per_image_token": 2.1875e-06, + "input_cost_per_token": 2.1875e-06, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 64000, @@ -404,9 +404,9 @@ }, "apac.amazon.nova-2-pro-preview-20251202-v1:0": { "cache_read_input_token_cost": 5.46875e-07, - "input_cost_per_token": 2.1875e-06, - "input_cost_per_image_token": 2.1875e-06, "input_cost_per_audio_token": 2.1875e-06, + "input_cost_per_image_token": 2.1875e-06, + "input_cost_per_token": 2.1875e-06, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 64000, @@ -440,9 +440,9 @@ }, "eu.amazon.nova-2-pro-preview-20251202-v1:0": { "cache_read_input_token_cost": 5.46875e-07, - "input_cost_per_token": 2.1875e-06, - "input_cost_per_image_token": 2.1875e-06, "input_cost_per_audio_token": 2.1875e-06, + "input_cost_per_image_token": 2.1875e-06, + "input_cost_per_token": 2.1875e-06, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 64000, @@ -476,9 +476,9 @@ }, "us.amazon.nova-2-pro-preview-20251202-v1:0": { "cache_read_input_token_cost": 5.46875e-07, - "input_cost_per_token": 2.1875e-06, - "input_cost_per_image_token": 2.1875e-06, "input_cost_per_audio_token": 2.1875e-06, + "input_cost_per_image_token": 2.1875e-06, + "input_cost_per_token": 2.1875e-06, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 64000, @@ -494,21 +494,21 @@ "supports_vision": true }, "amazon.nova-2-multimodal-embeddings-v1:0": { + "input_cost_per_audio_per_second": 0.00014, + "input_cost_per_image": 6e-05, + "input_cost_per_token": 1.35e-07, + "input_cost_per_video_per_second": 0.0007, "litellm_provider": "bedrock", "max_input_tokens": 8172, "max_tokens": 8172, "mode": "embedding", - "input_cost_per_token": 1.35e-07, - "input_cost_per_image": 6e-05, - "input_cost_per_video_per_second": 0.0007, - "input_cost_per_audio_per_second": 0.00014, "output_cost_per_token": 0.0, "output_vector_size": 3072, "source": "https://us-east-1.console.aws.amazon.com/bedrock/home?region=us-east-1#/model-catalog/serverless/amazon.nova-2-multimodal-embeddings-v1:0", + "supports_audio_input": true, "supports_embedding_image_input": true, "supports_image_input": true, - "supports_video_input": true, - "supports_audio_input": true + "supports_video_input": true }, "amazon.nova-micro-v1:0": { "input_cost_per_token": 3.5e-08, @@ -594,21 +594,21 @@ }, "amazon.titan-image-generator-v1": { "input_cost_per_image": 0.0, - "output_cost_per_image": 0.008, "litellm_provider": "bedrock", - "mode": "image_generation" + "mode": "image_generation", + "output_cost_per_image": 0.008 }, "amazon.titan-image-generator-v2": { "input_cost_per_image": 0.0, - "output_cost_per_image": 0.008, "litellm_provider": "bedrock", - "mode": "image_generation" + "mode": "image_generation", + "output_cost_per_image": 0.008 }, "amazon.titan-image-generator-v2:0": { "input_cost_per_image": 0.0, - "output_cost_per_image": 0.008, "litellm_provider": "bedrock", - "mode": "image_generation" + "mode": "image_generation", + "output_cost_per_image": 0.008 }, "twelvelabs.marengo-embed-2-7-v1:0": { "deprecation_date": "2026-11-30", @@ -624,10 +624,10 @@ }, "us.twelvelabs.marengo-embed-2-7-v1:0": { "deprecation_date": "2026-11-30", - "input_cost_per_token": 7e-05, - "input_cost_per_video_per_second": 0.0007, "input_cost_per_audio_per_second": 0.00014, "input_cost_per_image": 0.0001, + "input_cost_per_token": 7e-05, + "input_cost_per_video_per_second": 0.0007, "litellm_provider": "bedrock", "max_input_tokens": 77, "max_tokens": 77, @@ -639,10 +639,10 @@ }, "eu.twelvelabs.marengo-embed-2-7-v1:0": { "deprecation_date": "2026-11-30", - "input_cost_per_token": 7e-05, - "input_cost_per_video_per_second": 0.0007, "input_cost_per_audio_per_second": 0.00014, "input_cost_per_image": 0.0001, + "input_cost_per_token": 7e-05, + "input_cost_per_video_per_second": 0.0007, "litellm_provider": "bedrock", "max_input_tokens": 77, "max_tokens": 77, @@ -654,23 +654,23 @@ }, "twelvelabs.pegasus-1-2-v1:0": { "input_cost_per_video_per_second": 0.00049, - "output_cost_per_token": 7.5e-06, "litellm_provider": "bedrock", "mode": "chat", + "output_cost_per_token": 7.5e-06, "supports_video_input": true }, "us.twelvelabs.pegasus-1-2-v1:0": { "input_cost_per_video_per_second": 0.00049, - "output_cost_per_token": 7.5e-06, "litellm_provider": "bedrock", "mode": "chat", + "output_cost_per_token": 7.5e-06, "supports_video_input": true }, "eu.twelvelabs.pegasus-1-2-v1:0": { "input_cost_per_video_per_second": 0.00049, - "output_cost_per_token": 7.5e-06, "litellm_provider": "bedrock", "mode": "chat", + "output_cost_per_token": 7.5e-06, "supports_video_input": true }, "amazon.titan-text-express-v1": { @@ -710,41 +710,41 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 4e-06, + "prompt_cache_min_tokens": 2048, "supports_assistant_prefill": true, "supports_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_response_schema": true, - "supports_tool_choice": true, - "prompt_cache_min_tokens": 2048 + "supports_tool_choice": true }, "anthropic.claude-haiku-4-5-20251001-v1:0": { "cache_creation_input_token_cost": 1.25e-06, "cache_creation_input_token_cost_above_1hr": 2e-06, "cache_read_input_token_cost": 1e-07, "input_cost_per_token": 1e-06, + "input_cost_per_token_batches": 5e-07, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, + "output_cost_per_token_batches": 2.5e-06, + "prompt_cache_min_tokens": 4096, "source": "https://aws.amazon.com/about-aws/whats-new/2025/10/claude-4-5-haiku-anthropic-amazon-bedrock", "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096, - "input_cost_per_token_batches": 5e-07, - "output_cost_per_token_batches": 2.5e-06 + "supports_tool_search": true, + "supports_vision": true }, "anthropic.claude-haiku-4-5@20251001": { "cache_creation_input_token_cost": 1.25e-06, @@ -752,28 +752,30 @@ "cache_read_input_token_cost": 1e-07, "input_cost_per_token": 1e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, + "prompt_cache_min_tokens": 4096, "source": "https://aws.amazon.com/about-aws/whats-new/2025/10/claude-4-5-haiku-anthropic-amazon-bedrock", "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_streaming": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_streaming": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "anthropic.claude-3-5-sonnet-20240620-v1:0": { + "cache_creation_input_token_cost": 3.75e-06, + "cache_read_input_token_cost": 3e-07, "input_cost_per_token": 3e-06, "litellm_provider": "bedrock", "max_input_tokens": 200000, @@ -785,9 +787,7 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_creation_input_token_cost": 3.75e-06, - "cache_read_input_token_cost": 3e-07 + "supports_vision": true }, "anthropic.claude-3-5-sonnet-20241022-v2:0": { "cache_creation_input_token_cost": 3.75e-06, @@ -849,6 +849,8 @@ "supports_vision": true }, "anthropic.claude-3-haiku-20240307-v1:0": { + "cache_creation_input_token_cost": 3.125e-07, + "cache_read_input_token_cost": 2.5e-08, "deprecation_date": "2026-09-10", "input_cost_per_token": 2.5e-07, "litellm_provider": "bedrock", @@ -861,11 +863,11 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 2.5e-08, - "cache_creation_input_token_cost": 3.125e-07 + "supports_vision": true }, "anthropic.claude-3-opus-20240229-v1:0": { + "cache_creation_input_token_cost": 1.875e-05, + "cache_read_input_token_cost": 1.5e-06, "input_cost_per_token": 1.5e-05, "litellm_provider": "bedrock", "max_input_tokens": 200000, @@ -876,11 +878,11 @@ "supports_function_calling": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 1.5e-06, - "cache_creation_input_token_cost": 1.875e-05 + "supports_vision": true }, "anthropic.claude-3-sonnet-20240229-v1:0": { + "cache_creation_input_token_cost": 3.75e-06, + "cache_read_input_token_cost": 3e-07, "deprecation_date": "2026-07-30", "input_cost_per_token": 3e-06, "litellm_provider": "bedrock", @@ -893,9 +895,7 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-07, - "cache_creation_input_token_cost": 3.75e-06 + "supports_vision": true }, "anthropic.claude-instant-v1": { "input_cost_per_token": 8e-07, @@ -918,6 +918,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -931,8 +932,7 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "anthropic.claude-opus-4-20250514-v1:0": { "cache_creation_input_token_cost": 1.875e-05, @@ -944,6 +944,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -957,21 +958,21 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "anthropic.claude-opus-4-5-20251101-v1:0": { + "bedrock_output_config_effort_ceiling": "high", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -980,390 +981,395 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "high", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "anthropic.claude-opus-4-6-v1": { - "supports_adaptive_thinking": true, + "bedrock_output_config_effort_ceiling": "max", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "supports_max_reasoning_effort": true, - "bedrock_output_config_effort_ceiling": "max", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "global.anthropic.claude-opus-4-6-v1": { - "supports_adaptive_thinking": true, + "bedrock_output_config_effort_ceiling": "max", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "supports_max_reasoning_effort": true, - "bedrock_output_config_effort_ceiling": "max", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "us.anthropic.claude-opus-4-6-v1": { - "supports_adaptive_thinking": true, + "bedrock_output_config_effort_ceiling": "max", "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "supports_max_reasoning_effort": true, - "bedrock_output_config_effort_ceiling": "max", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "eu.anthropic.claude-opus-4-6-v1": { - "supports_adaptive_thinking": true, + "bedrock_output_config_effort_ceiling": "max", "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "supports_max_reasoning_effort": true, - "bedrock_output_config_effort_ceiling": "max", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "au.anthropic.claude-opus-4-6-v1": { - "supports_adaptive_thinking": true, + "bedrock_output_config_effort_ceiling": "max", "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "supports_max_reasoning_effort": true, - "bedrock_output_config_effort_ceiling": "max", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "anthropic.claude-opus-4-7": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 2048, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_sampling_params": false, "supports_tool_choice": true, + "supports_tool_search": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "anthropic.claude-mythos-preview": { "input_cost_per_token": 0, - "output_cost_per_token": 0, "litellm_provider": "bedrock", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", - "thinking_always_on": true, + "output_cost_per_token": 0, "supports_function_calling": true, - "supports_vision": true, + "supports_output_config": true, "supports_prompt_caching": false, "supports_reasoning": true, "supports_tool_choice": true, - "supports_output_config": true + "supports_vision": true, + "thinking_always_on": true }, "global.anthropic.claude-opus-4-7": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 2048, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_sampling_params": false, "supports_tool_choice": true, + "supports_tool_search": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "us.anthropic.claude-opus-4-7": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 2048, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_sampling_params": false, "supports_tool_choice": true, + "supports_tool_search": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "eu.anthropic.claude-opus-4-7": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 2048, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_sampling_params": false, "supports_tool_choice": true, + "supports_tool_search": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "au.anthropic.claude-opus-4-7": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 2048, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_sampling_params": false, "supports_tool_choice": true, + "supports_tool_search": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "anthropic.claude-fable-5": { + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 1.25e-05, "cache_creation_input_token_cost_above_1hr": 2e-05, "cache_read_input_token_cost": 1e-06, @@ -1374,17 +1380,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 5e-05, + "prompt_cache_min_tokens": 512, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "thinking_always_on": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1393,14 +1403,10 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 512 + "thinking_always_on": true }, "global.anthropic.claude-fable-5": { + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 1.25e-05, "cache_creation_input_token_cost_above_1hr": 2e-05, "cache_read_input_token_cost": 1e-06, @@ -1411,17 +1417,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 5e-05, + "prompt_cache_min_tokens": 512, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "thinking_always_on": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1430,14 +1440,10 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 512 + "thinking_always_on": true }, "us.anthropic.claude-fable-5": { + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 1.375e-05, "cache_creation_input_token_cost_above_1hr": 2.2e-05, "cache_read_input_token_cost": 1.1e-06, @@ -1448,17 +1454,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 5.5e-05, + "prompt_cache_min_tokens": 512, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "thinking_always_on": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1467,14 +1477,10 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 512 + "thinking_always_on": true }, "eu.anthropic.claude-fable-5": { + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 1.375e-05, "cache_creation_input_token_cost_above_1hr": 2.2e-05, "cache_read_input_token_cost": 1.1e-06, @@ -1485,17 +1491,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 5.5e-05, + "prompt_cache_min_tokens": 512, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "thinking_always_on": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1504,17 +1514,10 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 512 + "thinking_always_on": true }, "anthropic.claude-opus-5": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, @@ -1525,14 +1528,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 512, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1540,17 +1550,10 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 512 + "supports_xhigh_reasoning_effort": true }, "global.anthropic.claude-opus-5": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, @@ -1561,14 +1564,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 512, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1576,17 +1586,10 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 512 + "supports_xhigh_reasoning_effort": true }, "us.anthropic.claude-opus-5": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, @@ -1597,14 +1600,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 512, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1612,17 +1622,10 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 512 + "supports_xhigh_reasoning_effort": true }, "eu.anthropic.claude-opus-5": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, @@ -1633,14 +1636,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 512, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1648,17 +1658,10 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 512 + "supports_xhigh_reasoning_effort": true }, "au.anthropic.claude-opus-5": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, @@ -1669,14 +1672,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 512, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1684,17 +1694,10 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 512 + "supports_xhigh_reasoning_effort": true }, "jp.anthropic.claude-opus-5": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, @@ -1705,14 +1708,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 512, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1720,17 +1730,11 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 512 + "supports_xhigh_reasoning_effort": true }, "anthropic.claude-opus-4-8": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, @@ -1741,14 +1745,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1756,18 +1767,11 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "global.anthropic.claude-opus-4-8": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, @@ -1778,14 +1782,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1793,18 +1804,11 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "us.anthropic.claude-opus-4-8": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, @@ -1815,14 +1819,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1830,18 +1841,11 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "eu.anthropic.claude-opus-4-8": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, @@ -1852,14 +1856,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1867,18 +1878,11 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "au.anthropic.claude-opus-4-8": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, @@ -1889,14 +1893,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1904,18 +1915,11 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "jp.anthropic.claude-opus-4-8": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, @@ -1926,14 +1930,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1941,13 +1952,7 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "jp.anthropic.claude-opus-4-7": { "bedrock_converse_supports_strict_tools": false, @@ -1955,12 +1960,12 @@ "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 2048, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -1970,22 +1975,23 @@ "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_minimal_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_sampling_params": false, "supports_tool_choice": true, + "supports_tool_search": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_minimal_reasoning_effort": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "anthropic.claude-sonnet-5": { "bedrock_converse_supports_strict_tools": false, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 2.5e-06, "cache_creation_input_token_cost_above_1hr": 4e-06, "cache_read_input_token_cost": 2e-07, @@ -1996,16 +2002,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -2013,16 +2024,11 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "global.anthropic.claude-sonnet-5": { "bedrock_converse_supports_strict_tools": false, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 2.5e-06, "cache_creation_input_token_cost_above_1hr": 4e-06, "cache_read_input_token_cost": 2e-07, @@ -2033,16 +2039,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -2050,16 +2061,11 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "us.anthropic.claude-sonnet-5": { "bedrock_converse_supports_strict_tools": false, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 2.75e-06, "cache_creation_input_token_cost_above_1hr": 4.4e-06, "cache_read_input_token_cost": 2.2e-07, @@ -2070,16 +2076,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1.1e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -2087,16 +2098,11 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "eu.anthropic.claude-sonnet-5": { "bedrock_converse_supports_strict_tools": false, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 2.75e-06, "cache_creation_input_token_cost_above_1hr": 4.4e-06, "cache_read_input_token_cost": 2.2e-07, @@ -2107,16 +2113,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1.1e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -2124,16 +2135,11 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "au.anthropic.claude-sonnet-5": { "bedrock_converse_supports_strict_tools": false, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 2.75e-06, "cache_creation_input_token_cost_above_1hr": 4.4e-06, "cache_read_input_token_cost": 2.2e-07, @@ -2144,16 +2150,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1.1e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -2161,16 +2172,11 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "jp.anthropic.claude-sonnet-5": { "bedrock_converse_supports_strict_tools": false, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 2.75e-06, "cache_creation_input_token_cost_above_1hr": 4.4e-06, "cache_read_input_token_cost": 2.2e-07, @@ -2181,16 +2187,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1.1e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -2198,227 +2209,229 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "anthropic.claude-sonnet-4-6": { - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, "cache_read_input_token_cost": 3e-07, "input_cost_per_token": 3e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_tool_search": true, + "supports_vision": true }, "global.anthropic.claude-sonnet-4-6": { - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, "cache_read_input_token_cost": 3e-07, "input_cost_per_token": 3e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_tool_search": true, + "supports_vision": true }, "us.anthropic.claude-sonnet-4-6": { - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 4.125e-06, "cache_creation_input_token_cost_above_1hr": 6.6e-06, "cache_read_input_token_cost": 3.3e-07, "input_cost_per_token": 3.3e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.65e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_tool_search": true, + "supports_vision": true }, "eu.anthropic.claude-sonnet-4-6": { - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 4.125e-06, "cache_creation_input_token_cost_above_1hr": 6.6e-06, "cache_read_input_token_cost": 3.3e-07, "input_cost_per_token": 3.3e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.65e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_tool_search": true, + "supports_vision": true }, "au.anthropic.claude-sonnet-4-6": { - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 4.125e-06, "cache_creation_input_token_cost_above_1hr": 6.6e-06, "cache_read_input_token_cost": 3.3e-07, "input_cost_per_token": 3.3e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.65e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_tool_search": true, + "supports_vision": true }, "jp.anthropic.claude-sonnet-4-6": { - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 4.125e-06, "cache_creation_input_token_cost_above_1hr": 6.6e-06, "cache_read_input_token_cost": 3.3e-07, "input_cost_per_token": 3.3e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.65e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_tool_search": true, + "supports_vision": true }, "anthropic.claude-sonnet-4-20250514-v1:0": { + "bedrock_converse_supports_strict_tools": false, "cache_creation_input_token_cost": 3.75e-06, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, "deprecation_date": "2026-10-14", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -2432,27 +2445,27 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "bedrock_converse_supports_strict_tools": false, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "anthropic.claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, + "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.2e-05, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.2e-05, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "input_cost_per_token_batches": 1.5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "output_cost_per_token_batches": 7.5e-06, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -2461,17 +2474,15 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024, - "input_cost_per_token_batches": 1.5e-06, - "output_cost_per_token_batches": 7.5e-06 + "supports_tool_search": true, + "supports_vision": true }, "anthropic.claude-v1": { "input_cost_per_token": 8e-06, @@ -2651,6 +2662,8 @@ "supports_vision": true }, "apac.anthropic.claude-3-5-sonnet-20240620-v1:0": { + "cache_creation_input_token_cost": 3.75e-06, + "cache_read_input_token_cost": 3e-07, "deprecation_date": "2026-07-30", "input_cost_per_token": 3e-06, "litellm_provider": "bedrock", @@ -2663,9 +2676,7 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-07, - "cache_creation_input_token_cost": 3.75e-06 + "supports_vision": true }, "apac.anthropic.claude-3-5-sonnet-20241022-v2:0": { "cache_creation_input_token_cost": 3.75e-06, @@ -2688,6 +2699,8 @@ "supports_vision": true }, "apac.anthropic.claude-3-haiku-20240307-v1:0": { + "cache_creation_input_token_cost": 3.125e-07, + "cache_read_input_token_cost": 2.5e-08, "deprecation_date": "2026-09-10", "input_cost_per_token": 2.5e-07, "litellm_provider": "bedrock", @@ -2700,38 +2713,38 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 2.5e-08, - "cache_creation_input_token_cost": 3.125e-07 + "supports_vision": true }, "apac.anthropic.claude-haiku-4-5-20251001-v1:0": { "cache_creation_input_token_cost": 1.375e-06, "cache_read_input_token_cost": 1.1e-07, "input_cost_per_token": 1.1e-06, + "input_cost_per_token_batches": 5.5e-07, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5.5e-06, + "output_cost_per_token_batches": 2.75e-06, + "prompt_cache_min_tokens": 4096, "source": "https://aws.amazon.com/about-aws/whats-new/2025/10/claude-4-5-haiku-anthropic-amazon-bedrock", "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096, - "input_cost_per_token_batches": 5.5e-07, - "output_cost_per_token_batches": 2.75e-06 + "supports_tool_search": true, + "supports_vision": true }, "apac.anthropic.claude-3-sonnet-20240229-v1:0": { + "cache_creation_input_token_cost": 3.75e-06, + "cache_read_input_token_cost": 3e-07, "deprecation_date": "2026-07-30", "input_cost_per_token": 3e-06, "litellm_provider": "bedrock", @@ -2744,25 +2757,25 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-07, - "cache_creation_input_token_cost": 3.75e-06 + "supports_vision": true }, "apac.anthropic.claude-sonnet-4-20250514-v1:0": { + "bedrock_converse_supports_strict_tools": false, "cache_creation_input_token_cost": 3.75e-06, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, "deprecation_date": "2026-10-14", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -2776,9 +2789,7 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "bedrock_converse_supports_strict_tools": false, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "assemblyai/best": { "input_cost_per_second": 3.333e-05, @@ -2795,20 +2806,22 @@ "au.anthropic.claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 4.125e-06, "cache_creation_input_token_cost_above_1hr": 6.6e-06, + "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.32e-05, + "cache_creation_input_token_cost_above_200k_tokens": 8.25e-06, "cache_read_input_token_cost": 3.3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6.6e-07, "input_cost_per_token": 3.3e-06, "input_cost_per_token_above_200k_tokens": 6.6e-06, - "output_cost_per_token_above_200k_tokens": 2.475e-05, - "cache_creation_input_token_cost_above_200k_tokens": 8.25e-06, - "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.32e-05, - "cache_read_input_token_cost_above_200k_tokens": 6.6e-07, + "input_cost_per_token_batches": 1.65e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.65e-05, + "output_cost_per_token_above_200k_tokens": 2.475e-05, + "output_cost_per_token_batches": 8.25e-06, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -2817,17 +2830,15 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024, - "input_cost_per_token_batches": 1.65e-06, - "output_cost_per_token_batches": 8.25e-06 + "supports_tool_search": true, + "supports_vision": true }, "azure/ada": { "input_cost_per_token": 1e-07, @@ -2878,10 +2889,10 @@ "supports_function_calling": true }, "azure_ai/claude-haiku-4-5": { - "deprecation_date": "2026-10-19", "cache_creation_input_token_cost": 1.25e-06, "cache_creation_input_token_cost_above_1hr": 2e-06, "cache_read_input_token_cost": 1e-07, + "deprecation_date": "2026-10-19", "input_cost_per_token": 1e-06, "litellm_provider": "azure_ai", "max_input_tokens": 200000, @@ -2889,6 +2900,7 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, + "prompt_cache_min_tokens": 4096, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, @@ -2897,14 +2909,13 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "azure_ai/claude-opus-4-5": { - "deprecation_date": "2026-10-19", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2026-10-19", "input_cost_per_token": 5e-06, "litellm_provider": "azure_ai", "max_input_tokens": 200000, @@ -2912,70 +2923,73 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "azure_ai/claude-opus-4-6": { + "cache_creation_input_token_cost": 6.25e-06, + "cache_creation_input_token_cost_above_1hr": 1e-05, + "cache_read_input_token_cost": 5e-07, "deprecation_date": "2027-02-02", - "supports_adaptive_thinking": true, "input_cost_per_token": 5e-06, - "output_cost_per_token": 2.5e-05, "litellm_provider": "azure_ai", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, - "cache_creation_input_token_cost": 6.25e-06, - "cache_creation_input_token_cost_above_1hr": 1e-05, - "cache_read_input_token_cost": 5e-07, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "azure_ai/claude-opus-4-7": { + "cache_creation_input_token_cost": 6.25e-06, + "cache_creation_input_token_cost_above_1hr": 1e-05, + "cache_read_input_token_cost": 5e-07, "deprecation_date": "2027-04-06", - "supports_adaptive_thinking": true, "input_cost_per_token": 5e-06, - "output_cost_per_token": 2.5e-05, "litellm_provider": "azure_ai", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 2048, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, - "cache_creation_input_token_cost": 6.25e-06, - "cache_creation_input_token_cost_above_1hr": 1e-05, - "cache_read_input_token_cost": 5e-07, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -2983,32 +2997,31 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "azure_ai/claude-fable-5": { - "supports_mid_conversation_system": true, - "input_cost_per_token": 1e-05, - "output_cost_per_token": 5e-05, - "litellm_provider": "azure_ai", - "max_input_tokens": 1000000, - "max_output_tokens": 128000, - "max_tokens": 128000, - "mode": "chat", - "search_context_cost_per_query": { - "search_context_size_high": 0.01, - "search_context_size_low": 0.01, - "search_context_size_medium": 0.01 - }, "cache_creation_input_token_cost": 1.25e-05, "cache_creation_input_token_cost_above_1hr": 2e-05, "cache_read_input_token_cost": 1e-06, + "input_cost_per_token": 1e-05, + "litellm_provider": "azure_ai", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 5e-05, + "prompt_cache_min_tokens": 512, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, "supports_adaptive_thinking": true, - "thinking_always_on": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -3017,30 +3030,31 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 512 + "thinking_always_on": true }, "azure_ai/claude-opus-5": { - "supports_mid_conversation_system": true, - "supports_adaptive_thinking": true, + "cache_creation_input_token_cost": 6.25e-06, + "cache_creation_input_token_cost_above_1hr": 1e-05, + "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, - "output_cost_per_token": 2.5e-05, "litellm_provider": "azure_ai", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 512, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, - "cache_creation_input_token_cost": 6.25e-06, - "cache_creation_input_token_cost_above_1hr": 1e-05, - "cache_read_input_token_cost": 5e-07, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -3048,31 +3062,31 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 512 + "supports_xhigh_reasoning_effort": true }, "azure_ai/claude-opus-4-8": { - "supports_mid_conversation_system": true, - "supports_adaptive_thinking": true, + "cache_creation_input_token_cost": 6.25e-06, + "cache_creation_input_token_cost_above_1hr": 1e-05, + "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, - "output_cost_per_token": 2.5e-05, "litellm_provider": "azure_ai", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, - "cache_creation_input_token_cost": 6.25e-06, - "cache_creation_input_token_cost_above_1hr": 1e-05, - "cache_read_input_token_cost": 5e-07, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -3080,15 +3094,13 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "azure_ai/claude-opus-4-1": { - "deprecation_date": "2026-08-05", "cache_creation_input_token_cost": 1.875e-05, "cache_creation_input_token_cost_above_1hr": 3e-05, "cache_read_input_token_cost": 1.5e-06, + "deprecation_date": "2026-08-05", "input_cost_per_token": 1.5e-05, "litellm_provider": "azure_ai", "max_input_tokens": 200000, @@ -3096,6 +3108,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, @@ -3104,14 +3117,13 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "azure_ai/claude-sonnet-4-5": { - "deprecation_date": "2026-10-19", "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, "cache_read_input_token_cost": 3e-07, + "deprecation_date": "2026-10-19", "input_cost_per_token": 3e-06, "litellm_provider": "azure_ai", "max_input_tokens": 200000, @@ -3119,6 +3131,7 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, @@ -3127,11 +3140,9 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "azure_ai/claude-sonnet-5": { - "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 2.5e-06, "cache_creation_input_token_cost_above_1hr": 4e-06, "cache_read_input_token_cost": 2e-07, @@ -3142,6 +3153,7 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -3151,6 +3163,8 @@ "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -3158,16 +3172,13 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "azure_ai/claude-sonnet-4-6": { - "deprecation_date": "2027-02-10", - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, "cache_read_input_token_cost": 3e-07, + "deprecation_date": "2027-02-10", "input_cost_per_token": 3e-06, "litellm_provider": "azure_ai", "max_input_tokens": 1000000, @@ -3175,18 +3186,20 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "prompt_cache_min_tokens": 1024, + "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "azure/computer-use-preview": { "input_cost_per_token": 3e-06, @@ -3222,12 +3235,12 @@ }, "azure_ai/gpt-oss-120b": { "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, "litellm_provider": "azure_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 6e-07, "source": "https://azure.microsoft.com/en-us/pricing/details/cognitive-services/openai-service/", "supports_function_calling": true, "supports_parallel_function_calling": true, @@ -3235,15 +3248,15 @@ "supports_tool_choice": true }, "azure_ai/gpt-5.5": { - "deprecation_date": "2027-10-26", "cache_read_input_token_cost": 5e-07, "cache_read_input_token_cost_above_272k_tokens": 1e-06, - "cache_read_input_token_cost_priority": 1e-06, "cache_read_input_token_cost_above_272k_tokens_priority": 2e-06, + "cache_read_input_token_cost_priority": 1e-06, + "deprecation_date": "2027-10-26", "input_cost_per_token": 5e-06, "input_cost_per_token_above_272k_tokens": 1e-05, - "input_cost_per_token_priority": 1e-05, "input_cost_per_token_above_272k_tokens_priority": 2e-05, + "input_cost_per_token_priority": 1e-05, "litellm_provider": "azure_ai", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -3251,8 +3264,8 @@ "mode": "chat", "output_cost_per_token": 3e-05, "output_cost_per_token_above_272k_tokens": 4.5e-05, - "output_cost_per_token_priority": 6e-05, "output_cost_per_token_above_272k_tokens_priority": 9e-05, + "output_cost_per_token_priority": 6e-05, "source": "https://ai.azure.com/catalog/models/gpt-5.5", "supported_endpoints": [ "/v1/chat/completions", @@ -3267,7 +3280,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -3277,19 +3292,17 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure_ai/gpt-5.5-2026-04-23": { "cache_read_input_token_cost": 5e-07, "cache_read_input_token_cost_above_272k_tokens": 1e-06, - "cache_read_input_token_cost_priority": 1e-06, "cache_read_input_token_cost_above_272k_tokens_priority": 2e-06, + "cache_read_input_token_cost_priority": 1e-06, "input_cost_per_token": 5e-06, "input_cost_per_token_above_272k_tokens": 1e-05, - "input_cost_per_token_priority": 1e-05, "input_cost_per_token_above_272k_tokens_priority": 2e-05, + "input_cost_per_token_priority": 1e-05, "litellm_provider": "azure_ai", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -3297,8 +3310,8 @@ "mode": "chat", "output_cost_per_token": 3e-05, "output_cost_per_token_above_272k_tokens": 4.5e-05, - "output_cost_per_token_priority": 6e-05, "output_cost_per_token_above_272k_tokens_priority": 9e-05, + "output_cost_per_token_priority": 6e-05, "source": "https://ai.azure.com/catalog/models/gpt-5.5", "supported_endpoints": [ "/v1/chat/completions", @@ -3313,7 +3326,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -3323,20 +3338,18 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure_ai/gpt-5.4": { - "deprecation_date": "2027-09-02", "cache_read_input_token_cost": 2.5e-07, "cache_read_input_token_cost_above_272k_tokens": 5e-07, - "cache_read_input_token_cost_priority": 5e-07, "cache_read_input_token_cost_above_272k_tokens_priority": 1e-06, + "cache_read_input_token_cost_priority": 5e-07, + "deprecation_date": "2027-09-02", "input_cost_per_token": 2.5e-06, "input_cost_per_token_above_272k_tokens": 5e-06, - "input_cost_per_token_priority": 5e-06, "input_cost_per_token_above_272k_tokens_priority": 1e-05, + "input_cost_per_token_priority": 5e-06, "litellm_provider": "azure_ai", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -3344,8 +3357,8 @@ "mode": "chat", "output_cost_per_token": 1.5e-05, "output_cost_per_token_above_272k_tokens": 2.25e-05, - "output_cost_per_token_priority": 3e-05, "output_cost_per_token_above_272k_tokens_priority": 4.5e-05, + "output_cost_per_token_priority": 3e-05, "source": "https://ai.azure.com/catalog/models/gpt-5.4", "supported_endpoints": [ "/v1/chat/completions", @@ -3360,31 +3373,31 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true, - "supports_service_tier": true + "supports_xhigh_reasoning_effort": true }, "azure_ai/gpt-5.4-2026-03-05": { - "deprecation_date": "2027-09-02", "cache_read_input_token_cost": 2.5e-07, "cache_read_input_token_cost_above_272k_tokens": 5e-07, - "cache_read_input_token_cost_priority": 5e-07, "cache_read_input_token_cost_above_272k_tokens_priority": 1e-06, + "cache_read_input_token_cost_priority": 5e-07, + "deprecation_date": "2027-09-02", "input_cost_per_token": 2.5e-06, "input_cost_per_token_above_272k_tokens": 5e-06, - "input_cost_per_token_priority": 5e-06, "input_cost_per_token_above_272k_tokens_priority": 1e-05, + "input_cost_per_token_priority": 5e-06, "litellm_provider": "azure_ai", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -3392,8 +3405,8 @@ "mode": "chat", "output_cost_per_token": 1.5e-05, "output_cost_per_token_above_272k_tokens": 2.25e-05, - "output_cost_per_token_priority": 3e-05, "output_cost_per_token_above_272k_tokens_priority": 4.5e-05, + "output_cost_per_token_priority": 3e-05, "source": "https://ai.azure.com/catalog/models/gpt-5.4", "supported_endpoints": [ "/v1/chat/completions", @@ -3408,31 +3421,31 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true, - "supports_service_tier": true + "supports_xhigh_reasoning_effort": true }, "azure_ai/gpt-5.4-pro": { - "deprecation_date": "2027-09-07", "cache_read_input_token_cost": 3e-06, "cache_read_input_token_cost_above_272k_tokens": 6e-06, - "cache_read_input_token_cost_priority": 6e-06, "cache_read_input_token_cost_above_272k_tokens_priority": 1.2e-05, + "cache_read_input_token_cost_priority": 6e-06, + "deprecation_date": "2027-09-07", "input_cost_per_token": 3e-05, "input_cost_per_token_above_272k_tokens": 6e-05, - "input_cost_per_token_priority": 6e-05, "input_cost_per_token_above_272k_tokens_priority": 0.00012, + "input_cost_per_token_priority": 6e-05, "litellm_provider": "azure_ai", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -3440,8 +3453,8 @@ "mode": "responses", "output_cost_per_token": 0.00018, "output_cost_per_token_above_272k_tokens": 0.00027, - "output_cost_per_token_priority": 0.00036, "output_cost_per_token_above_272k_tokens_priority": 0.00054, + "output_cost_per_token_priority": 0.00036, "source": "https://ai.azure.com/catalog/models/gpt-5.4-pro", "supported_endpoints": [ "/v1/batch", @@ -3455,7 +3468,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -3465,20 +3480,18 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "azure_ai/gpt-5.4-pro-2026-03-05": { - "deprecation_date": "2027-09-07", "cache_read_input_token_cost": 3e-06, "cache_read_input_token_cost_above_272k_tokens": 6e-06, - "cache_read_input_token_cost_priority": 6e-06, "cache_read_input_token_cost_above_272k_tokens_priority": 1.2e-05, + "cache_read_input_token_cost_priority": 6e-06, + "deprecation_date": "2027-09-07", "input_cost_per_token": 3e-05, "input_cost_per_token_above_272k_tokens": 6e-05, - "input_cost_per_token_priority": 6e-05, "input_cost_per_token_above_272k_tokens_priority": 0.00012, + "input_cost_per_token_priority": 6e-05, "litellm_provider": "azure_ai", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -3486,8 +3499,8 @@ "mode": "responses", "output_cost_per_token": 0.00018, "output_cost_per_token_above_272k_tokens": 0.00027, - "output_cost_per_token_priority": 0.00036, "output_cost_per_token_above_272k_tokens_priority": 0.00054, + "output_cost_per_token_priority": 0.00036, "source": "https://ai.azure.com/catalog/models/gpt-5.4-pro", "supported_endpoints": [ "/v1/batch", @@ -3501,7 +3514,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -3511,14 +3526,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "azure_ai/gpt-5.4-mini": { - "deprecation_date": "2027-09-21", "cache_read_input_token_cost": 7.5e-08, "cache_read_input_token_cost_priority": 1.5e-07, + "deprecation_date": "2027-09-21", "input_cost_per_token": 7.5e-07, "input_cost_per_token_priority": 1.5e-06, "litellm_provider": "azure_ai", @@ -3542,25 +3555,25 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "supports_service_tier": true + "supports_xhigh_reasoning_effort": true }, "azure_ai/gpt-5.4-mini-2026-03-17": { - "deprecation_date": "2027-09-21", "cache_read_input_token_cost": 7.5e-08, "cache_read_input_token_cost_priority": 1.5e-07, + "deprecation_date": "2027-09-21", "input_cost_per_token": 7.5e-07, "input_cost_per_token_priority": 1.5e-06, "litellm_provider": "azure_ai", @@ -3584,25 +3597,25 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "supports_service_tier": true + "supports_xhigh_reasoning_effort": true }, "azure_ai/gpt-5.4-nano": { - "deprecation_date": "2027-09-21", "cache_read_input_token_cost": 2e-08, "cache_read_input_token_cost_priority": 4e-08, + "deprecation_date": "2027-09-21", "input_cost_per_token": 2e-07, "input_cost_per_token_priority": 4e-07, "litellm_provider": "azure_ai", @@ -3626,7 +3639,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -3636,14 +3651,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure_ai/gpt-5.4-nano-2026-03-17": { - "deprecation_date": "2027-09-21", "cache_read_input_token_cost": 2e-08, "cache_read_input_token_cost_priority": 4e-08, + "deprecation_date": "2027-09-21", "input_cost_per_token": 2e-07, "input_cost_per_token_priority": 4e-07, "litellm_provider": "azure_ai", @@ -3667,7 +3680,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -3677,21 +3692,19 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure_ai/model_router": { + "comment": "Flat cost of $0.14 per M input tokens for Azure AI Foundry Model Router infrastructure. Use pattern: azure_ai/model_router/ where deployment-name is your Azure deployment (e.g., azure-model-router)", "input_cost_per_token": 1.4e-07, - "output_cost_per_token": 0, "litellm_provider": "azure_ai", "mode": "chat", - "source": "https://azure.microsoft.com/en-us/pricing/details/ai-services/", - "comment": "Flat cost of $0.14 per M input tokens for Azure AI Foundry Model Router infrastructure. Use pattern: azure_ai/model_router/ where deployment-name is your Azure deployment (e.g., azure-model-router)" + "output_cost_per_token": 0, + "source": "https://azure.microsoft.com/en-us/pricing/details/ai-services/" }, "azure/eu/gpt-4o-2024-08-06": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 1.375e-06, + "deprecation_date": "2027-04-14", "input_cost_per_token": 2.75e-06, "litellm_provider": "azure", "max_input_tokens": 128000, @@ -3707,8 +3720,9 @@ "supports_vision": true }, "azure/eu/gpt-4o-2024-11-20": { - "deprecation_date": "2027-04-14", "cache_creation_input_token_cost": 1.38e-06, + "cache_read_input_token_cost": 1.375e-06, + "deprecation_date": "2027-04-14", "input_cost_per_token": 2.75e-06, "litellm_provider": "azure", "max_input_tokens": 128000, @@ -3720,8 +3734,7 @@ "supports_parallel_function_calling": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 1.375e-06 + "supports_vision": true }, "azure/eu/gpt-4o-mini-2024-07-18": { "cache_read_input_token_cost": 8.3e-08, @@ -3872,8 +3885,8 @@ "supports_vision": true }, "azure/eu/gpt-5.1": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 1.4e-07, + "deprecation_date": "2027-05-15", "input_cost_per_token": 1.38e-06, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -3896,6 +3909,7 @@ ], "supports_function_calling": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -3903,8 +3917,7 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_none_reasoning_effort": true + "supports_vision": true }, "azure/eu/gpt-5.1-chat": { "cache_read_input_token_cost": 1.4e-07, @@ -3931,6 +3944,7 @@ ], "supports_function_calling": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -3938,12 +3952,11 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_none_reasoning_effort": true + "supports_vision": true }, "azure/eu/gpt-5.1-codex": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 1.4e-07, + "deprecation_date": "2027-05-15", "input_cost_per_token": 1.38e-06, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -3973,8 +3986,8 @@ "supports_vision": true }, "azure/eu/gpt-5.1-codex-mini": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 2.8e-08, + "deprecation_date": "2027-05-15", "input_cost_per_token": 2.75e-07, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -4148,8 +4161,8 @@ "supports_vision": true }, "azure/global/gpt-4o-2024-08-06": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 1.25e-06, + "deprecation_date": "2027-04-14", "input_cost_per_token": 2.5e-06, "litellm_provider": "azure", "max_input_tokens": 128000, @@ -4165,8 +4178,8 @@ "supports_vision": true }, "azure/global/gpt-4o-2024-11-20": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 1.25e-06, + "deprecation_date": "2027-04-14", "input_cost_per_token": 2.5e-06, "litellm_provider": "azure", "max_input_tokens": 128000, @@ -4182,8 +4195,8 @@ "supports_vision": true }, "azure/global/gpt-5.1": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 1.25e-07, + "deprecation_date": "2027-05-15", "input_cost_per_token": 1.25e-06, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -4206,6 +4219,7 @@ ], "supports_function_calling": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -4213,8 +4227,7 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_none_reasoning_effort": true + "supports_vision": true }, "azure/global/gpt-5.1-chat": { "cache_read_input_token_cost": 1.25e-07, @@ -4241,6 +4254,7 @@ ], "supports_function_calling": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -4248,12 +4262,11 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_none_reasoning_effort": true + "supports_vision": true }, "azure/global/gpt-5.1-codex": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 1.25e-07, + "deprecation_date": "2027-05-15", "input_cost_per_token": 1.25e-06, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -4283,8 +4296,8 @@ "supports_vision": true }, "azure/global/gpt-5.1-codex-mini": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 2.5e-08, + "deprecation_date": "2027-05-15", "input_cost_per_token": 2.5e-07, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -4522,8 +4535,8 @@ "supports_vision": true }, "azure/gpt-4.1": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-04-14", "input_cost_per_token": 2e-06, "input_cost_per_token_batches": 1e-06, "litellm_provider": "azure", @@ -4556,8 +4569,8 @@ "supports_web_search": false }, "azure/gpt-4.1-2025-04-14": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-04-14", "input_cost_per_token": 2e-06, "input_cost_per_token_batches": 1e-06, "litellm_provider": "azure", @@ -4590,8 +4603,8 @@ "supports_web_search": false }, "azure/gpt-4.1-mini": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 1e-07, + "deprecation_date": "2027-04-14", "input_cost_per_token": 4e-07, "input_cost_per_token_batches": 2e-07, "litellm_provider": "azure", @@ -4624,8 +4637,8 @@ "supports_web_search": false }, "azure/gpt-4.1-mini-2025-04-14": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 1e-07, + "deprecation_date": "2027-04-14", "input_cost_per_token": 4e-07, "input_cost_per_token_batches": 2e-07, "litellm_provider": "azure", @@ -4658,8 +4671,8 @@ "supports_web_search": false }, "azure/gpt-4.1-nano": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 2.5e-08, + "deprecation_date": "2027-04-14", "input_cost_per_token": 1e-07, "input_cost_per_token_batches": 5e-08, "litellm_provider": "azure", @@ -4691,8 +4704,8 @@ "supports_vision": true }, "azure/gpt-4.1-nano-2025-04-14": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 2.5e-08, + "deprecation_date": "2027-04-14", "input_cost_per_token": 1e-07, "input_cost_per_token_batches": 5e-08, "litellm_provider": "azure", @@ -4774,8 +4787,8 @@ "supports_vision": true }, "azure/gpt-4o-2024-08-06": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 1.25e-06, + "deprecation_date": "2027-04-14", "input_cost_per_token": 2.5e-06, "litellm_provider": "azure", "max_input_tokens": 128000, @@ -4791,8 +4804,8 @@ "supports_vision": true }, "azure/gpt-4o-2024-11-20": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 1.25e-06, + "deprecation_date": "2027-04-14", "input_cost_per_token": 2.75e-06, "litellm_provider": "azure", "max_input_tokens": 128000, @@ -4861,6 +4874,8 @@ "text", "audio" ], + "supports_audio_input": true, + "supports_audio_output": true, "supports_function_calling": true, "supports_native_streaming": true, "supports_parallel_function_calling": true, @@ -4869,9 +4884,7 @@ "supports_response_schema": false, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": false, - "supports_audio_input": true, - "supports_audio_output": true + "supports_vision": false }, "azure/gpt-audio-mini": { "deprecation_date": "2027-04-06", @@ -4895,6 +4908,8 @@ "text", "audio" ], + "supports_audio_input": true, + "supports_audio_output": true, "supports_function_calling": true, "supports_native_streaming": true, "supports_parallel_function_calling": true, @@ -4903,9 +4918,7 @@ "supports_response_schema": false, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": false, - "supports_audio_input": true, - "supports_audio_output": true + "supports_vision": false }, "azure/gpt-audio-mini-2025-10-06": { "deprecation_date": "2027-04-06", @@ -4929,6 +4942,8 @@ "text", "audio" ], + "supports_audio_input": true, + "supports_audio_output": true, "supports_function_calling": true, "supports_native_streaming": true, "supports_parallel_function_calling": true, @@ -4937,9 +4952,7 @@ "supports_response_schema": false, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": false, - "supports_audio_input": true, - "supports_audio_output": true + "supports_vision": false }, "azure/gpt-4o-audio-preview-2024-12-17": { "input_cost_per_audio_token": 4e-05, @@ -4973,8 +4986,8 @@ "supports_vision": false }, "azure/gpt-4o-mini": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 7.5e-08, + "deprecation_date": "2027-04-14", "input_cost_per_token": 1.65e-07, "litellm_provider": "azure", "max_input_tokens": 128000, @@ -5061,6 +5074,7 @@ "cache_read_input_token_cost": 4e-07, "deprecation_date": "2027-03-02", "input_cost_per_audio_token": 3.2e-05, + "input_cost_per_image_token": 5e-06, "input_cost_per_token": 4e-06, "litellm_provider": "azure", "max_input_tokens": 32000, @@ -5086,14 +5100,14 @@ "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "input_cost_per_image_token": 5e-06 + "supports_tool_choice": true }, "azure/gpt-realtime-1.5-2026-02-23": { "cache_creation_input_audio_token_cost": 4e-07, "cache_read_input_token_cost": 4e-07, "deprecation_date": "2027-08-24", "input_cost_per_audio_token": 3.2e-05, + "input_cost_per_image_token": 5e-06, "input_cost_per_token": 4e-06, "litellm_provider": "azure", "max_input_tokens": 32000, @@ -5119,13 +5133,13 @@ "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "input_cost_per_image_token": 5e-06 + "supports_tool_choice": true }, "azure/gpt-realtime-mini": { "cache_creation_input_audio_token_cost": 3e-07, "cache_read_input_token_cost": 6e-08, "input_cost_per_audio_token": 1e-05, + "input_cost_per_image_token": 8e-07, "input_cost_per_token": 6e-07, "litellm_provider": "azure", "max_input_tokens": 32000, @@ -5151,13 +5165,13 @@ "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "input_cost_per_image_token": 8e-07 + "supports_tool_choice": true }, "azure/gpt-realtime-mini-2025-10-06": { "cache_creation_input_audio_token_cost": 3e-07, "cache_read_input_token_cost": 6e-08, "input_cost_per_audio_token": 1e-05, + "input_cost_per_image_token": 8e-07, "input_cost_per_token": 6e-07, "litellm_provider": "azure", "max_input_tokens": 32000, @@ -5183,8 +5197,7 @@ "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "input_cost_per_image_token": 8e-07 + "supports_tool_choice": true }, "azure/gpt-4o-mini-transcribe": { "input_cost_per_audio_token": 1.25e-06, @@ -5330,7 +5343,9 @@ "image" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -5338,9 +5353,7 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_none_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_vision": true }, "azure/gpt-5.1-chat-2025-11-13": { "cache_read_input_token_cost": 1.25e-07, @@ -5369,6 +5382,7 @@ ], "supports_function_calling": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": false, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -5376,8 +5390,7 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": false, - "supports_vision": true, - "supports_none_reasoning_effort": true + "supports_vision": true }, "azure/gpt-5.1-codex-2025-11-13": { "cache_read_input_token_cost": 1.25e-07, @@ -5448,8 +5461,8 @@ "supports_vision": true }, "azure/gpt-5": { - "deprecation_date": "2027-02-09", "cache_read_input_token_cost": 1.25e-07, + "deprecation_date": "2027-02-09", "input_cost_per_token": 1.25e-06, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -5612,8 +5625,8 @@ "supports_vision": true }, "azure/gpt-5-mini": { - "deprecation_date": "2027-02-09", "cache_read_input_token_cost": 2.5e-08, + "deprecation_date": "2027-02-09", "input_cost_per_token": 2.5e-07, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -5678,8 +5691,8 @@ "supports_vision": true }, "azure/gpt-5-nano": { - "deprecation_date": "2027-02-09", "cache_read_input_token_cost": 5e-09, + "deprecation_date": "2027-02-09", "input_cost_per_token": 5e-08, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -5774,8 +5787,8 @@ "supports_vision": true }, "azure/gpt-5.1": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 1.25e-07, + "deprecation_date": "2027-05-15", "input_cost_per_token": 1.25e-06, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -5798,6 +5811,7 @@ ], "supports_function_calling": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -5805,8 +5819,7 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_none_reasoning_effort": true + "supports_vision": true }, "azure/gpt-5.1-chat": { "cache_read_input_token_cost": 1.25e-07, @@ -5833,6 +5846,7 @@ ], "supports_function_calling": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -5840,12 +5854,11 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_none_reasoning_effort": true + "supports_vision": true }, "azure/gpt-5.1-codex": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 1.25e-07, + "deprecation_date": "2027-05-15", "input_cost_per_token": 1.25e-06, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -5906,8 +5919,8 @@ "supports_vision": true }, "azure/gpt-5.1-codex-mini": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 2.5e-08, + "deprecation_date": "2027-05-15", "input_cost_per_token": 2.5e-07, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -5937,8 +5950,8 @@ "supports_vision": true }, "azure/gpt-5.2": { - "deprecation_date": "2027-06-08", "cache_read_input_token_cost": 1.75e-07, + "deprecation_date": "2027-06-08", "input_cost_per_token": 1.75e-06, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -6247,15 +6260,15 @@ "supports_web_search": true }, "azure/gpt-5.4": { - "deprecation_date": "2027-09-02", "cache_read_input_token_cost": 2.5e-07, "cache_read_input_token_cost_above_272k_tokens": 5e-07, - "cache_read_input_token_cost_priority": 5e-07, "cache_read_input_token_cost_above_272k_tokens_priority": 1e-06, + "cache_read_input_token_cost_priority": 5e-07, + "deprecation_date": "2027-09-02", "input_cost_per_token": 2.5e-06, "input_cost_per_token_above_272k_tokens": 5e-06, - "input_cost_per_token_priority": 5e-06, "input_cost_per_token_above_272k_tokens_priority": 1e-05, + "input_cost_per_token_priority": 5e-06, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -6263,8 +6276,8 @@ "mode": "chat", "output_cost_per_token": 1.5e-05, "output_cost_per_token_above_272k_tokens": 2.25e-05, - "output_cost_per_token_priority": 3e-05, "output_cost_per_token_above_272k_tokens_priority": 4.5e-05, + "output_cost_per_token_priority": 3e-05, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6278,7 +6291,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6287,23 +6302,21 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "azure/us/gpt-5.4": { - "deprecation_date": "2027-09-02", "cache_read_input_token_cost": 2.8e-07, "cache_read_input_token_cost_priority": 5.5e-07, + "deprecation_date": "2027-09-02", "input_cost_per_token": 2.75e-06, "input_cost_per_token_priority": 5.5e-06, - "output_cost_per_token": 1.65e-05, - "output_cost_per_token_priority": 3.3e-05, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1.65e-05, + "output_cost_per_token_priority": 3.3e-05, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6317,7 +6330,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6326,23 +6341,21 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "azure/eu/gpt-5.4": { - "deprecation_date": "2027-09-02", "cache_read_input_token_cost": 2.8e-07, "cache_read_input_token_cost_priority": 5.5e-07, + "deprecation_date": "2027-09-02", "input_cost_per_token": 2.75e-06, "input_cost_per_token_priority": 5.5e-06, - "output_cost_per_token": 1.65e-05, - "output_cost_per_token_priority": 3.3e-05, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1.65e-05, + "output_cost_per_token_priority": 3.3e-05, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6356,7 +6369,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6365,20 +6380,18 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "azure/gpt-5.4-2026-03-05": { "cache_read_input_token_cost": 2.5e-07, "cache_read_input_token_cost_above_272k_tokens": 5e-07, - "cache_read_input_token_cost_priority": 5e-07, "cache_read_input_token_cost_above_272k_tokens_priority": 1e-06, + "cache_read_input_token_cost_priority": 5e-07, "deprecation_date": "2027-09-02", "input_cost_per_token": 2.5e-06, "input_cost_per_token_above_272k_tokens": 5e-06, - "input_cost_per_token_priority": 5e-06, "input_cost_per_token_above_272k_tokens_priority": 1e-05, + "input_cost_per_token_priority": 5e-06, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -6386,8 +6399,8 @@ "mode": "chat", "output_cost_per_token": 1.5e-05, "output_cost_per_token_above_272k_tokens": 2.25e-05, - "output_cost_per_token_priority": 3e-05, "output_cost_per_token_above_272k_tokens_priority": 4.5e-05, + "output_cost_per_token_priority": 3e-05, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6401,7 +6414,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6410,9 +6425,7 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "azure/us/gpt-5.4-2026-03-05": { "cache_read_input_token_cost": 2.8e-07, @@ -6420,13 +6433,13 @@ "deprecation_date": "2027-09-02", "input_cost_per_token": 2.75e-06, "input_cost_per_token_priority": 5.5e-06, - "output_cost_per_token": 1.65e-05, - "output_cost_per_token_priority": 3.3e-05, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1.65e-05, + "output_cost_per_token_priority": 3.3e-05, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6440,7 +6453,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6449,9 +6464,7 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "azure/eu/gpt-5.4-2026-03-05": { "cache_read_input_token_cost": 2.8e-07, @@ -6459,13 +6472,13 @@ "deprecation_date": "2027-09-02", "input_cost_per_token": 2.75e-06, "input_cost_per_token_priority": 5.5e-06, - "output_cost_per_token": 1.65e-05, - "output_cost_per_token_priority": 3.3e-05, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1.65e-05, + "output_cost_per_token_priority": 3.3e-05, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6479,7 +6492,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6488,14 +6503,12 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "azure/gpt-5.4-pro": { - "deprecation_date": "2027-09-07", "cache_read_input_token_cost": 3e-06, "cache_read_input_token_cost_above_272k_tokens": 6e-06, + "deprecation_date": "2027-09-07", "input_cost_per_token": 3e-05, "input_cost_per_token_above_272k_tokens": 6e-05, "litellm_provider": "azure", @@ -6575,6 +6588,9 @@ "supports_web_search": true }, "azure/gpt-5.6": { + "cache_creation_input_token_cost": 6.25e-06, + "cache_creation_input_token_cost_above_272k_tokens": 1.25e-05, + "cache_creation_input_token_cost_priority": 1.25e-05, "cache_read_input_token_cost": 5e-07, "cache_read_input_token_cost_above_272k_tokens": 1e-06, "cache_read_input_token_cost_priority": 1e-06, @@ -6590,6 +6606,7 @@ "output_cost_per_token_above_272k_tokens": 4.5e-05, "output_cost_per_token_priority": 6e-05, "search_context_cost_per_query": {}, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6603,7 +6620,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6613,15 +6632,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, + "supports_xhigh_reasoning_effort": true + }, + "azure/gpt-5.6-sol": { "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_272k_tokens": 1.25e-05, "cache_creation_input_token_cost_priority": 1.25e-05, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service" - }, - "azure/gpt-5.6-sol": { "cache_read_input_token_cost": 5e-07, "cache_read_input_token_cost_above_272k_tokens": 1e-06, "cache_read_input_token_cost_priority": 1e-06, @@ -6642,6 +6658,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6655,7 +6672,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6665,15 +6684,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "cache_creation_input_token_cost": 6.25e-06, - "cache_creation_input_token_cost_above_272k_tokens": 1.25e-05, - "cache_creation_input_token_cost_priority": 1.25e-05, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service" + "supports_xhigh_reasoning_effort": true }, "azure/gpt-5.6-terra": { + "cache_creation_input_token_cost": 2.5e-06, + "cache_creation_input_token_cost_above_272k_tokens": 5e-06, + "cache_creation_input_token_cost_priority": 5e-06, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_272k_tokens": 4e-07, "cache_read_input_token_cost_priority": 4e-07, @@ -6694,6 +6710,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6707,7 +6724,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6717,15 +6736,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "cache_creation_input_token_cost": 2.5e-06, - "cache_creation_input_token_cost_above_272k_tokens": 5e-06, - "cache_creation_input_token_cost_priority": 5e-06, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service" + "supports_xhigh_reasoning_effort": true }, "azure/gpt-5.6-luna": { + "cache_creation_input_token_cost": 2.5e-07, + "cache_creation_input_token_cost_above_272k_tokens": 5e-07, + "cache_creation_input_token_cost_priority": 5e-07, "cache_read_input_token_cost": 2e-08, "cache_read_input_token_cost_above_272k_tokens": 4e-08, "cache_read_input_token_cost_priority": 4e-08, @@ -6746,6 +6762,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6759,7 +6776,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6769,15 +6788,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "cache_creation_input_token_cost": 2.5e-07, - "cache_creation_input_token_cost_above_272k_tokens": 5e-07, - "cache_creation_input_token_cost_priority": 5e-07, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service" + "supports_xhigh_reasoning_effort": true }, "azure/us/gpt-5.6": { + "cache_creation_input_token_cost": 6.875e-06, + "cache_creation_input_token_cost_above_272k_tokens": 1.375e-05, + "cache_creation_input_token_cost_priority": 1.375e-05, "cache_read_input_token_cost": 5.5e-07, "cache_read_input_token_cost_above_272k_tokens": 1.1e-06, "cache_read_input_token_cost_priority": 1.1e-06, @@ -6797,6 +6813,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6810,7 +6827,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6820,15 +6839,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "cache_creation_input_token_cost": 6.875e-06, - "cache_creation_input_token_cost_above_272k_tokens": 1.375e-05, - "cache_creation_input_token_cost_priority": 1.375e-05, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service" + "supports_xhigh_reasoning_effort": true }, "azure/us/gpt-5.6-sol": { + "cache_creation_input_token_cost": 6.875e-06, + "cache_creation_input_token_cost_above_272k_tokens": 1.375e-05, + "cache_creation_input_token_cost_priority": 1.375e-05, "cache_read_input_token_cost": 5.5e-07, "cache_read_input_token_cost_above_272k_tokens": 1.1e-06, "cache_read_input_token_cost_priority": 1.1e-06, @@ -6849,6 +6865,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6862,7 +6879,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6872,15 +6891,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "cache_creation_input_token_cost": 6.875e-06, - "cache_creation_input_token_cost_above_272k_tokens": 1.375e-05, - "cache_creation_input_token_cost_priority": 1.375e-05, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service" + "supports_xhigh_reasoning_effort": true }, "azure/us/gpt-5.6-terra": { + "cache_creation_input_token_cost": 2.75e-06, + "cache_creation_input_token_cost_above_272k_tokens": 5.5e-06, + "cache_creation_input_token_cost_priority": 5.5e-06, "cache_read_input_token_cost": 2.2e-07, "cache_read_input_token_cost_above_272k_tokens": 4.4e-07, "cache_read_input_token_cost_priority": 4.4e-07, @@ -6901,6 +6917,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6914,7 +6931,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6924,15 +6943,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "cache_creation_input_token_cost_above_272k_tokens": 5.5e-06, - "cache_creation_input_token_cost": 2.75e-06, - "cache_creation_input_token_cost_priority": 5.5e-06, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service" + "supports_xhigh_reasoning_effort": true }, "azure/us/gpt-5.6-luna": { + "cache_creation_input_token_cost": 2.75e-07, + "cache_creation_input_token_cost_above_272k_tokens": 5.5e-07, + "cache_creation_input_token_cost_priority": 5.5e-07, "cache_read_input_token_cost": 2.2e-08, "cache_read_input_token_cost_above_272k_tokens": 4.4e-08, "cache_read_input_token_cost_priority": 4.4e-08, @@ -6953,6 +6969,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6966,7 +6983,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6976,15 +6995,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "cache_creation_input_token_cost": 2.75e-07, - "cache_creation_input_token_cost_above_272k_tokens": 5.5e-07, - "cache_creation_input_token_cost_priority": 5.5e-07, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service" + "supports_xhigh_reasoning_effort": true }, "azure/eu/gpt-5.6": { + "cache_creation_input_token_cost": 6.875e-06, + "cache_creation_input_token_cost_above_272k_tokens": 1.375e-05, + "cache_creation_input_token_cost_priority": 1.375e-05, "cache_read_input_token_cost": 5.5e-07, "cache_read_input_token_cost_above_272k_tokens": 1.1e-06, "cache_read_input_token_cost_priority": 1.1e-06, @@ -7004,6 +7020,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -7017,7 +7034,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7027,15 +7046,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, + "supports_xhigh_reasoning_effort": true + }, + "azure/eu/gpt-5.6-sol": { "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_272k_tokens": 1.375e-05, "cache_creation_input_token_cost_priority": 1.375e-05, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service" - }, - "azure/eu/gpt-5.6-sol": { "cache_read_input_token_cost": 5.5e-07, "cache_read_input_token_cost_above_272k_tokens": 1.1e-06, "cache_read_input_token_cost_priority": 1.1e-06, @@ -7056,6 +7072,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -7069,7 +7086,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7079,15 +7098,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "cache_creation_input_token_cost": 6.875e-06, - "cache_creation_input_token_cost_above_272k_tokens": 1.375e-05, - "cache_creation_input_token_cost_priority": 1.375e-05, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service" + "supports_xhigh_reasoning_effort": true }, "azure/eu/gpt-5.6-terra": { + "cache_creation_input_token_cost": 2.75e-06, + "cache_creation_input_token_cost_above_272k_tokens": 5.5e-06, + "cache_creation_input_token_cost_priority": 5.5e-06, "cache_read_input_token_cost": 2.2e-07, "cache_read_input_token_cost_above_272k_tokens": 4.4e-07, "cache_read_input_token_cost_priority": 4.4e-07, @@ -7108,6 +7124,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -7121,7 +7138,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7131,15 +7150,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "cache_creation_input_token_cost_above_272k_tokens": 5.5e-06, - "cache_creation_input_token_cost_priority": 5.5e-06, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", - "cache_creation_input_token_cost": 2.75e-06 + "supports_xhigh_reasoning_effort": true }, "azure/eu/gpt-5.6-luna": { + "cache_creation_input_token_cost": 2.75e-07, + "cache_creation_input_token_cost_above_272k_tokens": 5.5e-07, + "cache_creation_input_token_cost_priority": 5.5e-07, "cache_read_input_token_cost": 2.2e-08, "cache_read_input_token_cost_above_272k_tokens": 4.4e-08, "cache_read_input_token_cost_priority": 4.4e-08, @@ -7160,6 +7176,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -7173,7 +7190,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7183,24 +7202,18 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "cache_creation_input_token_cost": 2.75e-07, - "cache_creation_input_token_cost_above_272k_tokens": 5.5e-07, - "cache_creation_input_token_cost_priority": 5.5e-07, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service" + "supports_xhigh_reasoning_effort": true }, "azure/gpt-5.5": { - "deprecation_date": "2027-10-26", "cache_read_input_token_cost": 5e-07, "cache_read_input_token_cost_above_272k_tokens": 1e-06, - "cache_read_input_token_cost_priority": 1e-06, "cache_read_input_token_cost_above_272k_tokens_priority": 2e-06, + "cache_read_input_token_cost_priority": 1e-06, + "deprecation_date": "2027-10-26", "input_cost_per_token": 5e-06, "input_cost_per_token_above_272k_tokens": 1e-05, - "input_cost_per_token_priority": 1e-05, "input_cost_per_token_above_272k_tokens_priority": 2e-05, + "input_cost_per_token_priority": 1e-05, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -7208,8 +7221,8 @@ "mode": "chat", "output_cost_per_token": 3e-05, "output_cost_per_token_above_272k_tokens": 4.5e-05, - "output_cost_per_token_priority": 6e-05, "output_cost_per_token_above_272k_tokens_priority": 9e-05, + "output_cost_per_token_priority": 6e-05, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -7228,7 +7241,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7238,26 +7253,24 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure/us/gpt-5.5": { - "deprecation_date": "2027-10-26", "cache_read_input_token_cost": 5.5e-07, "cache_read_input_token_cost_above_272k_tokens": 1.1e-06, "cache_read_input_token_cost_priority": 1.38e-06, + "deprecation_date": "2027-10-26", "input_cost_per_token": 5.5e-06, "input_cost_per_token_above_272k_tokens": 1.1e-05, "input_cost_per_token_priority": 1.375e-05, - "output_cost_per_token": 3.3e-05, - "output_cost_per_token_above_272k_tokens": 4.95e-05, - "output_cost_per_token_priority": 8.25e-05, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.3e-05, + "output_cost_per_token_above_272k_tokens": 4.95e-05, + "output_cost_per_token_priority": 8.25e-05, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -7276,7 +7289,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7286,26 +7301,24 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure/eu/gpt-5.5": { - "deprecation_date": "2027-10-26", "cache_read_input_token_cost": 5.5e-07, "cache_read_input_token_cost_above_272k_tokens": 1.1e-06, "cache_read_input_token_cost_priority": 1.38e-06, + "deprecation_date": "2027-10-26", "input_cost_per_token": 5.5e-06, "input_cost_per_token_above_272k_tokens": 1.1e-05, "input_cost_per_token_priority": 1.375e-05, - "output_cost_per_token": 3.3e-05, - "output_cost_per_token_above_272k_tokens": 4.95e-05, - "output_cost_per_token_priority": 8.25e-05, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.3e-05, + "output_cost_per_token_above_272k_tokens": 4.95e-05, + "output_cost_per_token_priority": 8.25e-05, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -7324,7 +7337,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7334,19 +7349,17 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure/gpt-5.5-2026-04-23": { "cache_read_input_token_cost": 5e-07, "cache_read_input_token_cost_above_272k_tokens": 1e-06, - "cache_read_input_token_cost_priority": 1e-06, "cache_read_input_token_cost_above_272k_tokens_priority": 2e-06, + "cache_read_input_token_cost_priority": 1e-06, "input_cost_per_token": 5e-06, "input_cost_per_token_above_272k_tokens": 1e-05, - "input_cost_per_token_priority": 1e-05, "input_cost_per_token_above_272k_tokens_priority": 2e-05, + "input_cost_per_token_priority": 1e-05, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -7354,8 +7367,8 @@ "mode": "chat", "output_cost_per_token": 3e-05, "output_cost_per_token_above_272k_tokens": 4.5e-05, - "output_cost_per_token_priority": 6e-05, "output_cost_per_token_above_272k_tokens_priority": 9e-05, + "output_cost_per_token_priority": 6e-05, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -7392,14 +7405,14 @@ "input_cost_per_token": 5.5e-06, "input_cost_per_token_above_272k_tokens": 1.1e-05, "input_cost_per_token_priority": 1.375e-05, - "output_cost_per_token": 3.3e-05, - "output_cost_per_token_above_272k_tokens": 4.95e-05, - "output_cost_per_token_priority": 8.25e-05, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.3e-05, + "output_cost_per_token_above_272k_tokens": 4.95e-05, + "output_cost_per_token_priority": 8.25e-05, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -7436,14 +7449,14 @@ "input_cost_per_token": 5.5e-06, "input_cost_per_token_above_272k_tokens": 1.1e-05, "input_cost_per_token_priority": 1.375e-05, - "output_cost_per_token": 3.3e-05, - "output_cost_per_token_above_272k_tokens": 4.95e-05, - "output_cost_per_token_priority": 8.25e-05, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.3e-05, + "output_cost_per_token_above_272k_tokens": 4.95e-05, + "output_cost_per_token_priority": 8.25e-05, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -7502,7 +7515,10 @@ "text" ], "supports_function_calling": true, + "supports_low_reasoning_effort": false, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7512,10 +7528,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "supports_low_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure/gpt-5.5-pro-2026-04-23": { "cache_read_input_token_cost": 3e-06, @@ -7558,8 +7571,8 @@ "supports_web_search": true }, "azure/gpt-5.4-mini": { - "deprecation_date": "2027-09-21", "cache_read_input_token_cost": 7.5e-08, + "deprecation_date": "2027-09-21", "input_cost_per_token": 7.5e-07, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -7585,7 +7598,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7595,9 +7610,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure/gpt-5.4-mini-2026-03-17": { "cache_read_input_token_cost": 7.5e-08, @@ -7627,7 +7640,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7637,13 +7652,11 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure/gpt-5.4-nano": { - "deprecation_date": "2027-09-21", "cache_read_input_token_cost": 2e-08, + "deprecation_date": "2027-09-21", "input_cost_per_token": 2e-07, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -7669,7 +7682,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7679,9 +7694,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure/gpt-5.4-nano-2026-03-17": { "cache_read_input_token_cost": 2e-08, @@ -7711,7 +7724,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7721,9 +7736,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure/gpt-image-1": { "cache_read_input_token_cost": 1.25e-06, @@ -7860,64 +7873,64 @@ ] }, "azure/gpt-image-1.5": { - "deprecation_date": "2027-06-16", "cache_read_input_token_cost": 1.25e-06, - "input_cost_per_token": 5e-06, + "deprecation_date": "2027-06-16", "input_cost_per_image_token": 8e-06, + "input_cost_per_token": 5e-06, "litellm_provider": "azure", "mode": "image_generation", "output_cost_per_image_token": 3.2e-05, + "output_cost_per_token": 1e-05, "supported_endpoints": [ "/v1/images/generations", "/v1/images/edits" - ], - "output_cost_per_token": 1e-05 + ] }, "azure/gpt-image-1.5-2025-12-16": { "cache_read_input_token_cost": 1.25e-06, "deprecation_date": "2027-06-16", - "input_cost_per_token": 5e-06, "input_cost_per_image_token": 8e-06, + "input_cost_per_token": 5e-06, "litellm_provider": "azure", "mode": "image_generation", "output_cost_per_image_token": 3.2e-05, + "output_cost_per_token": 1e-05, "supported_endpoints": [ "/v1/images/generations", "/v1/images/edits" - ], - "output_cost_per_token": 1e-05 + ] }, "azure/gpt-image-2": { - "deprecation_date": "2027-10-21", "cache_read_input_token_cost": 1.25e-06, - "input_cost_per_token": 5e-06, + "deprecation_date": "2027-10-21", "input_cost_per_image_token": 8e-06, + "input_cost_per_token": 5e-06, "litellm_provider": "azure", "mode": "image_generation", - "output_cost_per_token": 1e-05, "output_cost_per_image_token": 3e-05, + "output_cost_per_token": 1e-05, "supported_endpoints": [ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "azure/gpt-image-2-2026-04-21": { "cache_read_input_token_cost": 1.25e-06, "deprecation_date": "2027-10-21", - "input_cost_per_token": 5e-06, "input_cost_per_image_token": 8e-06, + "input_cost_per_token": 5e-06, "litellm_provider": "azure", "mode": "image_generation", - "output_cost_per_token": 1e-05, "output_cost_per_image_token": 3e-05, + "output_cost_per_token": 1e-05, "supported_endpoints": [ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "azure/low/1024-x-1024/gpt-image-1-mini": { "deprecation_date": "2027-04-07", @@ -8028,8 +8041,8 @@ "supports_function_calling": true }, "azure/o1": { - "deprecation_date": "2026-10-21", "cache_read_input_token_cost": 7.5e-06, + "deprecation_date": "2026-10-21", "input_cost_per_token": 1.5e-05, "litellm_provider": "azure", "max_input_tokens": 200000, @@ -8123,8 +8136,8 @@ "supports_vision": false }, "azure/o3": { - "deprecation_date": "2026-10-21", "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2026-10-21", "input_cost_per_token": 2e-06, "litellm_provider": "azure", "max_input_tokens": 200000, @@ -8153,8 +8166,8 @@ "supports_vision": true }, "azure/o3-2025-04-16": { - "deprecation_date": "2026-10-21", "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2026-10-21", "input_cost_per_token": 2e-06, "litellm_provider": "azure", "max_input_tokens": 200000, @@ -8221,8 +8234,8 @@ "supports_web_search": true }, "azure/o3-mini": { - "deprecation_date": "2026-10-01", "cache_read_input_token_cost": 5.5e-07, + "deprecation_date": "2026-10-01", "input_cost_per_token": 1.1e-06, "litellm_provider": "azure", "max_input_tokens": 200000, @@ -8314,8 +8327,8 @@ "supports_vision": true }, "azure/o4-mini": { - "deprecation_date": "2026-10-16", "cache_read_input_token_cost": 2.75e-07, + "deprecation_date": "2026-10-16", "input_cost_per_token": 1.1e-06, "litellm_provider": "azure", "max_input_tokens": 200000, @@ -8448,8 +8461,8 @@ "mode": "audio_speech" }, "azure/us/gpt-4.1-2025-04-14": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 5.5e-07, + "deprecation_date": "2027-04-14", "input_cost_per_token": 2.2e-06, "input_cost_per_token_batches": 1.1e-06, "litellm_provider": "azure", @@ -8482,8 +8495,8 @@ "supports_web_search": false }, "azure/us/gpt-4.1-mini-2025-04-14": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 1.1e-07, + "deprecation_date": "2027-04-14", "input_cost_per_token": 4.4e-07, "input_cost_per_token_batches": 2.2e-07, "litellm_provider": "azure", @@ -8516,8 +8529,8 @@ "supports_web_search": false }, "azure/us/gpt-4.1-nano-2025-04-14": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 2.5e-08, + "deprecation_date": "2027-04-14", "input_cost_per_token": 1.1e-07, "input_cost_per_token_batches": 6e-08, "litellm_provider": "azure", @@ -8549,8 +8562,8 @@ "supports_vision": true }, "azure/us/gpt-4o-2024-08-06": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 1.375e-06, + "deprecation_date": "2027-04-14", "input_cost_per_token": 2.75e-06, "litellm_provider": "azure", "max_input_tokens": 128000, @@ -8566,8 +8579,9 @@ "supports_vision": true }, "azure/us/gpt-4o-2024-11-20": { - "deprecation_date": "2027-04-14", "cache_creation_input_token_cost": 1.38e-06, + "cache_read_input_token_cost": 1.375e-06, + "deprecation_date": "2027-04-14", "input_cost_per_token": 2.75e-06, "litellm_provider": "azure", "max_input_tokens": 128000, @@ -8579,8 +8593,7 @@ "supports_parallel_function_calling": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 1.375e-06 + "supports_vision": true }, "azure/us/gpt-4o-mini-2024-07-18": { "cache_read_input_token_cost": 8.3e-08, @@ -8764,8 +8777,8 @@ "supports_vision": true }, "azure/us/gpt-5.1": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 1.4e-07, + "deprecation_date": "2027-05-15", "input_cost_per_token": 1.38e-06, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -8788,6 +8801,7 @@ ], "supports_function_calling": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -8795,8 +8809,7 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_none_reasoning_effort": true + "supports_vision": true }, "azure/us/gpt-5.1-chat": { "cache_read_input_token_cost": 1.4e-07, @@ -8823,6 +8836,7 @@ ], "supports_function_calling": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -8830,12 +8844,11 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_none_reasoning_effort": true + "supports_vision": true }, "azure/us/gpt-5.1-codex": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 1.4e-07, + "deprecation_date": "2027-05-15", "input_cost_per_token": 1.38e-06, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -8865,8 +8878,8 @@ "supports_vision": true }, "azure/us/gpt-5.1-codex-mini": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 2.8e-08, + "deprecation_date": "2027-05-15", "input_cost_per_token": 2.75e-07, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -8942,8 +8955,8 @@ "supports_vision": false }, "azure/us/o3-2025-04-16": { - "deprecation_date": "2026-10-21", "cache_read_input_token_cost": 5.5e-07, + "deprecation_date": "2026-10-21", "input_cost_per_token": 2.2e-06, "litellm_provider": "azure", "max_input_tokens": 200000, @@ -9063,8 +9076,8 @@ ] }, "azure_ai/FW-DeepSeek-V3.2": { - "deprecation_date": "2027-07-01", "cache_read_input_token_cost": 3.1e-07, + "deprecation_date": "2027-07-01", "input_cost_per_token": 6.2e-07, "litellm_provider": "azure_ai", "max_input_tokens": 163840, @@ -9094,8 +9107,8 @@ "supports_tool_choice": true }, "azure_ai/FW-GLM-5": { - "deprecation_date": "2027-07-01", "cache_read_input_token_cost": 2.2e-07, + "deprecation_date": "2027-07-01", "input_cost_per_token": 1.1e-06, "litellm_provider": "azure_ai", "max_input_tokens": 200000, @@ -9110,8 +9123,8 @@ "supports_tool_choice": true }, "azure_ai/FW-GLM-5.1": { - "deprecation_date": "2027-07-01", "cache_read_input_token_cost": 2.86e-07, + "deprecation_date": "2027-07-01", "input_cost_per_token": 1.54e-06, "litellm_provider": "azure_ai", "max_input_tokens": 202800, @@ -9177,8 +9190,8 @@ "supports_tool_choice": true }, "azure_ai/FW-Kimi-K2.5": { - "deprecation_date": "2027-07-01", "cache_read_input_token_cost": 1.1e-07, + "deprecation_date": "2027-07-01", "input_cost_per_token": 6.6e-07, "litellm_provider": "azure_ai", "max_input_tokens": 262144, @@ -9270,8 +9283,8 @@ "supports_vision": true }, "azure_ai/FW-MiniMax-M2.5": { - "deprecation_date": "2027-07-01", "cache_read_input_token_cost": 3.3e-08, + "deprecation_date": "2027-07-01", "input_cost_per_token": 3.3e-07, "litellm_provider": "azure_ai", "max_input_tokens": 1000000, @@ -9645,54 +9658,54 @@ "output_cost_per_token": 5e-07, "source": "https://azure.microsoft.com/en-us/pricing/details/ai-foundry-models/microsoft/", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_reasoning": true + "supports_reasoning": true, + "supports_tool_choice": true }, "azure_ai/mistral-document-ai-2505": { "deprecation_date": "2026-07-20", "litellm_provider": "azure_ai", - "ocr_cost_per_page": 0.003, "mode": "ocr", + "ocr_cost_per_page": 0.003, + "source": "https://devblogs.microsoft.com/foundry/whats-new-in-azure-ai-foundry-august-2025/#mistral-document-ai-(ocr)-%E2%80%94-serverless-in-foundry", "supported_endpoints": [ "/v1/ocr" - ], - "source": "https://devblogs.microsoft.com/foundry/whats-new-in-azure-ai-foundry-august-2025/#mistral-document-ai-(ocr)-%E2%80%94-serverless-in-foundry" + ] }, "azure_ai/mistral-document-ai-2512": { "litellm_provider": "azure_ai", - "ocr_cost_per_page": 0.003, "mode": "ocr", + "ocr_cost_per_page": 0.003, + "source": "https://azure.microsoft.com/en-us/pricing/details/ai-foundry-models/", "supported_endpoints": [ "/v1/ocr" - ], - "source": "https://azure.microsoft.com/en-us/pricing/details/ai-foundry-models/" + ] }, "azure_ai/doc-intelligence/prebuilt-read": { "litellm_provider": "azure_ai", - "ocr_cost_per_page": 0.0015, "mode": "ocr", + "ocr_cost_per_page": 0.0015, + "source": "https://azure.microsoft.com/en-us/pricing/details/ai-document-intelligence/", "supported_endpoints": [ "/v1/ocr" - ], - "source": "https://azure.microsoft.com/en-us/pricing/details/ai-document-intelligence/" + ] }, "azure_ai/doc-intelligence/prebuilt-layout": { "litellm_provider": "azure_ai", - "ocr_cost_per_page": 0.01, "mode": "ocr", + "ocr_cost_per_page": 0.01, + "source": "https://azure.microsoft.com/en-us/pricing/details/ai-document-intelligence/", "supported_endpoints": [ "/v1/ocr" - ], - "source": "https://azure.microsoft.com/en-us/pricing/details/ai-document-intelligence/" + ] }, "azure_ai/doc-intelligence/prebuilt-document": { "litellm_provider": "azure_ai", - "ocr_cost_per_page": 0.01, "mode": "ocr", + "ocr_cost_per_page": 0.01, + "source": "https://azure.microsoft.com/en-us/pricing/details/ai-document-intelligence/", "supported_endpoints": [ "/v1/ocr" - ], - "source": "https://azure.microsoft.com/en-us/pricing/details/ai-document-intelligence/" + ] }, "azure_ai/MAI-DS-R1": { "input_cost_per_token": 1.35e-06, @@ -9983,12 +9996,12 @@ "azure_ai/grok-4-fast-non-reasoning": { "deprecation_date": "2026-05-01", "input_cost_per_token": 2e-07, - "output_cost_per_token": 5e-07, "litellm_provider": "azure_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 5e-07, "supports_function_calling": true, "supports_response_schema": true, "supports_tool_choice": true, @@ -9997,12 +10010,12 @@ "azure_ai/grok-4-fast-reasoning": { "deprecation_date": "2026-05-01", "input_cost_per_token": 2e-07, - "output_cost_per_token": 5e-07, "litellm_provider": "azure_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 5e-07, "source": "https://azure.microsoft.com/en-us/pricing/details/ai-foundry-models/grok/", "supports_function_calling": true, "supports_response_schema": true, @@ -10011,12 +10024,12 @@ }, "azure_ai/grok-4-1-fast-non-reasoning": { "input_cost_per_token": 2e-07, - "output_cost_per_token": 5e-07, "litellm_provider": "azure_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 5e-07, "source": "https://techcommunity.microsoft.com/t5/Azure-AI-Foundry-Blog/Grok-4-0-Goes-GA-in-Microsoft-Foundry-and-Grok-4-1-Fast-Arrives/ba-p/4497964", "supports_function_calling": true, "supports_response_schema": true, @@ -10025,12 +10038,12 @@ }, "azure_ai/grok-4-1-fast-reasoning": { "input_cost_per_token": 2e-07, - "output_cost_per_token": 5e-07, "litellm_provider": "azure_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 5e-07, "source": "https://techcommunity.microsoft.com/t5/Azure-AI-Foundry-Blog/Grok-4-0-Goes-GA-in-Microsoft-Foundry-and-Grok-4-1-Fast-Arrives/ba-p/4497964", "supports_function_calling": true, "supports_reasoning": true, @@ -10376,11 +10389,11 @@ "max_tokens": 163840, "mode": "chat", "output_cost_per_token": 2.22e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_reasoning": true, + "supports_tool_choice": true }, "bedrock/ap-northeast-1/minimax.minimax-m2.1": { "input_cost_per_token": 3.6e-07, @@ -10390,10 +10403,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/ap-northeast-1/minimax.minimax-m2.5": { "input_cost_per_token": 3.6e-07, @@ -10402,12 +10415,12 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.44e-06, "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "output_cost_per_token": 1.44e-06 + "supports_tool_choice": true }, "bedrock/ap-northeast-1/moonshotai.kimi-k2-thinking": { "input_cost_per_token": 7.3e-07, @@ -10428,11 +10441,11 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 3.6e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_vision": true }, "bedrock/ap-northeast-1/qwen.qwen3-coder-next": { "input_cost_per_token": 6e-07, @@ -10442,10 +10455,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/moonshotai.kimi-k2-thinking": { "input_cost_per_token": 7.3e-07, @@ -10500,11 +10513,11 @@ "max_tokens": 163840, "mode": "chat", "output_cost_per_token": 2.22e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_reasoning": true, + "supports_tool_choice": true }, "bedrock/ap-south-1/minimax.minimax-m2.1": { "input_cost_per_token": 3.6e-07, @@ -10514,10 +10527,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/ap-south-1/minimax.minimax-m2.5": { "input_cost_per_token": 3.6e-07, @@ -10526,12 +10539,12 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.44e-06, "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "output_cost_per_token": 1.44e-06 + "supports_tool_choice": true }, "bedrock/ap-south-1/moonshotai.kimi-k2-thinking": { "input_cost_per_token": 7.1e-07, @@ -10552,11 +10565,11 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 3.6e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_vision": true }, "bedrock/ap-south-1/qwen.qwen3-coder-next": { "input_cost_per_token": 6e-07, @@ -10566,10 +10579,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/ap-southeast-2/minimax.minimax-m2.5": { "input_cost_per_token": 3.09e-07, @@ -10578,12 +10591,12 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.236e-06, "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "output_cost_per_token": 1.236e-06 + "supports_tool_choice": true }, "bedrock/ap-southeast-3/deepseek.v3.2": { "input_cost_per_token": 7.4e-07, @@ -10593,11 +10606,11 @@ "max_tokens": 163840, "mode": "chat", "output_cost_per_token": 2.22e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_reasoning": true, + "supports_tool_choice": true }, "bedrock/ap-southeast-3/minimax.minimax-m2.1": { "input_cost_per_token": 3.6e-07, @@ -10607,10 +10620,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/ap-southeast-3/minimax.minimax-m2.5": { "input_cost_per_token": 3.6e-07, @@ -10619,12 +10632,12 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.44e-06, "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "output_cost_per_token": 1.44e-06 + "supports_tool_choice": true }, "bedrock/ap-southeast-3/moonshotai.kimi-k2.5": { "input_cost_per_token": 7.2e-07, @@ -10634,11 +10647,11 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 3.6e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_vision": true }, "bedrock/ap-southeast-3/qwen.qwen3-coder-next": { "input_cost_per_token": 6e-07, @@ -10648,10 +10661,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/ca-central-1/meta.llama3-70b-instruct-v1:0": { "input_cost_per_token": 3.05e-06, @@ -10679,11 +10692,11 @@ "max_tokens": 163840, "mode": "chat", "output_cost_per_token": 2.22e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_reasoning": true, + "supports_tool_choice": true }, "bedrock/eu-north-1/minimax.minimax-m2.1": { "input_cost_per_token": 3.6e-07, @@ -10693,10 +10706,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/eu-north-1/minimax.minimax-m2.5": { "input_cost_per_token": 3.6e-07, @@ -10705,12 +10718,12 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.44e-06, "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "output_cost_per_token": 1.44e-06 + "supports_tool_choice": true }, "bedrock/eu-north-1/moonshotai.kimi-k2.5": { "input_cost_per_token": 7.2e-07, @@ -10720,11 +10733,11 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 3.6e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_vision": true }, "bedrock/eu-central-1/1-month-commitment/anthropic.claude-instant-v1": { "input_cost_per_second": 0.01635, @@ -10821,10 +10834,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/eu-central-1/minimax.minimax-m2.5": { "input_cost_per_token": 3.6e-07, @@ -10833,12 +10846,12 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.44e-06, "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "output_cost_per_token": 1.44e-06 + "supports_tool_choice": true }, "bedrock/eu-central-1/qwen.qwen3-coder-next": { "input_cost_per_token": 6e-07, @@ -10848,10 +10861,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/eu-west-1/meta.llama3-70b-instruct-v1:0": { "input_cost_per_token": 2.86e-06, @@ -10879,10 +10892,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/eu-west-1/minimax.minimax-m2.5": { "input_cost_per_token": 3.6e-07, @@ -10891,12 +10904,12 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.44e-06, "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "output_cost_per_token": 1.44e-06 + "supports_tool_choice": true }, "bedrock/eu-west-1/qwen.qwen3-coder-next": { "input_cost_per_token": 6e-07, @@ -10906,10 +10919,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/eu-west-2/meta.llama3-70b-instruct-v1:0": { "input_cost_per_token": 3.45e-06, @@ -10937,10 +10950,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.86e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/eu-west-2/minimax.minimax-m2.5": { "input_cost_per_token": 4.7e-07, @@ -10949,12 +10962,12 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.86e-06, "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "output_cost_per_token": 1.86e-06 + "supports_tool_choice": true }, "bedrock/eu-west-2/qwen.qwen3-coder-next": { "input_cost_per_token": 7.8e-07, @@ -10964,10 +10977,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.86e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/eu-west-3/mistral.mistral-7b-instruct-v0:2": { "input_cost_per_token": 2e-07, @@ -11007,10 +11020,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/eu-south-1/minimax.minimax-m2.5": { "input_cost_per_token": 3.6e-07, @@ -11019,12 +11032,12 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.44e-06, "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "output_cost_per_token": 1.44e-06 + "supports_tool_choice": true }, "bedrock/eu-south-1/qwen.qwen3-coder-next": { "input_cost_per_token": 6e-07, @@ -11034,12 +11047,14 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/invoke/anthropic.claude-3-5-sonnet-20240620-v1:0": { + "cache_creation_input_token_cost": 3.75e-06, + "cache_read_input_token_cost": 3e-07, "input_cost_per_token": 3e-06, "litellm_provider": "bedrock", "max_input_tokens": 200000, @@ -11053,9 +11068,7 @@ "supports_function_calling": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-07, - "cache_creation_input_token_cost": 3.75e-06 + "supports_vision": true }, "bedrock/sa-east-1/meta.llama3-70b-instruct-v1:0": { "input_cost_per_token": 4.45e-06, @@ -11083,11 +11096,11 @@ "max_tokens": 163840, "mode": "chat", "output_cost_per_token": 2.22e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_reasoning": true, + "supports_tool_choice": true }, "bedrock/sa-east-1/minimax.minimax-m2.1": { "input_cost_per_token": 3.6e-07, @@ -11097,10 +11110,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/sa-east-1/minimax.minimax-m2.5": { "input_cost_per_token": 3.6e-07, @@ -11109,12 +11122,12 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.44e-06, "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "output_cost_per_token": 1.44e-06 + "supports_tool_choice": true }, "bedrock/sa-east-1/moonshotai.kimi-k2-thinking": { "input_cost_per_token": 7.3e-07, @@ -11135,11 +11148,11 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 3.6e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_vision": true }, "bedrock/sa-east-1/qwen.qwen3-coder-next": { "input_cost_per_token": 6e-07, @@ -11149,10 +11162,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/us-east-1/1-month-commitment/anthropic.claude-instant-v1": { "input_cost_per_second": 0.011, @@ -11298,11 +11311,11 @@ "max_tokens": 163840, "mode": "chat", "output_cost_per_token": 1.85e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_reasoning": true, + "supports_tool_choice": true }, "bedrock/us-east-1/minimax.minimax-m2.1": { "input_cost_per_token": 3e-07, @@ -11312,23 +11325,23 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/us-east-1/minimax.minimax-m2.5": { "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, "litellm_provider": "bedrock", "max_input_tokens": 1000000, "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/us-east-1/moonshotai.kimi-k2-thinking": { "input_cost_per_token": 6e-07, @@ -11349,11 +11362,11 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 3e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_vision": true }, "bedrock/us-east-1/qwen.qwen3-coder-next": { "input_cost_per_token": 5e-07, @@ -11363,10 +11376,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/us-east-2/deepseek.v3.2": { "input_cost_per_token": 6.2e-07, @@ -11376,11 +11389,11 @@ "max_tokens": 163840, "mode": "chat", "output_cost_per_token": 1.85e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_reasoning": true, + "supports_tool_choice": true }, "bedrock/us-east-2/minimax.minimax-m2.1": { "input_cost_per_token": 3e-07, @@ -11390,10 +11403,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/us-east-2/minimax.minimax-m2.5": { "input_cost_per_token": 3e-07, @@ -11402,12 +11415,12 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.2e-06, "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "output_cost_per_token": 1.2e-06 + "supports_tool_choice": true }, "bedrock/us-east-2/moonshotai.kimi-k2-thinking": { "input_cost_per_token": 6e-07, @@ -11428,11 +11441,11 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 3e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_vision": true }, "bedrock/us-east-2/qwen.qwen3-coder-next": { "input_cost_per_token": 5e-07, @@ -11442,10 +11455,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/us-gov-east-1/amazon.nova-pro-v1:0": { "input_cost_per_token": 9.6e-07, @@ -11507,6 +11520,8 @@ "output_cost_per_token": 1.5e-06 }, "bedrock/us-gov-east-1/anthropic.claude-3-5-sonnet-20240620-v1:0": { + "cache_creation_input_token_cost": 4.5e-06, + "cache_read_input_token_cost": 3.6e-07, "deprecation_date": "2026-07-30", "input_cost_per_token": 3.6e-06, "litellm_provider": "bedrock", @@ -11519,11 +11534,11 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3.6e-07, - "cache_creation_input_token_cost": 4.5e-06 + "supports_vision": true }, "bedrock/us-gov-east-1/anthropic.claude-3-haiku-20240307-v1:0": { + "cache_creation_input_token_cost": 3.75e-07, + "cache_read_input_token_cost": 3e-08, "deprecation_date": "2026-09-10", "input_cost_per_token": 3e-07, "litellm_provider": "bedrock", @@ -11536,9 +11551,7 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-08, - "cache_creation_input_token_cost": 3.75e-07 + "supports_vision": true }, "bedrock/us-gov-east-1/anthropic.claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 4.5e-06, @@ -11546,24 +11559,24 @@ "cache_read_input_token_cost": 3.6e-07, "input_cost_per_token": 3.6e-06, "litellm_provider": "bedrock", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.8e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_tool_search": true, + "supports_vision": true }, "bedrock/us-gov-east-1/claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 4.5e-06, @@ -11571,24 +11584,24 @@ "cache_read_input_token_cost": 3.6e-07, "input_cost_per_token": 3.6e-06, "litellm_provider": "bedrock", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.8e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_tool_search": true, + "supports_vision": true }, "bedrock/us-gov-east-1/meta.llama3-70b-instruct-v1:0": { "input_cost_per_token": 2.65e-06, @@ -11691,6 +11704,8 @@ "supports_vision": true }, "bedrock/us-gov-west-1/anthropic.claude-3-5-sonnet-20240620-v1:0": { + "cache_creation_input_token_cost": 4.5e-06, + "cache_read_input_token_cost": 3.6e-07, "deprecation_date": "2026-07-30", "input_cost_per_token": 3.6e-06, "litellm_provider": "bedrock", @@ -11703,11 +11718,11 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3.6e-07, - "cache_creation_input_token_cost": 4.5e-06 + "supports_vision": true }, "bedrock/us-gov-west-1/anthropic.claude-3-haiku-20240307-v1:0": { + "cache_creation_input_token_cost": 3.75e-07, + "cache_read_input_token_cost": 3e-08, "deprecation_date": "2026-09-10", "input_cost_per_token": 3e-07, "litellm_provider": "bedrock", @@ -11720,9 +11735,7 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-08, - "cache_creation_input_token_cost": 3.75e-07 + "supports_vision": true }, "bedrock/us-gov-west-1/anthropic.claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 4.5e-06, @@ -11730,24 +11743,24 @@ "cache_read_input_token_cost": 3.6e-07, "input_cost_per_token": 3.6e-06, "litellm_provider": "bedrock", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.8e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_tool_search": true, + "supports_vision": true }, "bedrock/us-gov-west-1/claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 4.5e-06, @@ -11755,24 +11768,24 @@ "cache_read_input_token_cost": 3.6e-07, "input_cost_per_token": 3.6e-06, "litellm_provider": "bedrock", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.8e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_tool_search": true, + "supports_vision": true }, "bedrock/us-gov-west-1/meta.llama3-70b-instruct-v1:0": { "input_cost_per_token": 2.65e-06, @@ -11938,11 +11951,11 @@ "max_tokens": 163840, "mode": "chat", "output_cost_per_token": 1.85e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_reasoning": true, + "supports_tool_choice": true }, "bedrock/us-west-2/minimax.minimax-m2.1": { "input_cost_per_token": 3e-07, @@ -11952,23 +11965,23 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/us-west-2/minimax.minimax-m2.5": { "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, "litellm_provider": "bedrock", "max_input_tokens": 1000000, "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/us-west-2/moonshotai.kimi-k2-thinking": { "input_cost_per_token": 6e-07, @@ -11989,11 +12002,11 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 3e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_vision": true }, "bedrock/us-west-2/qwen.qwen3-coder-next": { "input_cost_per_token": 5e-07, @@ -12003,10 +12016,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/us.anthropic.claude-3-5-haiku-20241022-v1:0": { "cache_creation_input_token_cost": 1e-06, @@ -12018,13 +12031,13 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 4e-06, + "prompt_cache_min_tokens": 2048, "supports_assistant_prefill": true, "supports_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_response_schema": true, - "supports_tool_choice": true, - "prompt_cache_min_tokens": 2048 + "supports_tool_choice": true }, "black_forest_labs/flux-kontext-pro": { "litellm_provider": "black_forest_labs", @@ -12176,6 +12189,7 @@ "supports_tool_choice": true }, "cerebras/zai-glm-4.7": { + "deprecation_date": "2026-08-17", "input_cost_per_token": 2.25e-06, "litellm_provider": "cerebras", "max_input_tokens": 128000, @@ -12185,10 +12199,9 @@ "output_cost_per_token": 2.75e-06, "source": "https://www.cerebras.ai/pricing", "supports_function_calling": true, + "supports_none_reasoning_effort": true, "supports_reasoning": true, - "supports_tool_choice": true, - "deprecation_date": "2026-08-17", - "supports_none_reasoning_effort": true + "supports_tool_choice": true }, "chatdolphin": { "input_cost_per_token": 5e-07, @@ -12229,10 +12242,10 @@ ] }, "claude-haiku-4-5-20251001": { - "deprecation_date": "2026-10-15", "cache_creation_input_token_cost": 1.25e-06, "cache_creation_input_token_cost_above_1hr": 2e-06, "cache_read_input_token_cost": 1e-07, + "deprecation_date": "2026-10-15", "input_cost_per_token": 1e-06, "litellm_provider": "anthropic", "max_input_tokens": 200000, @@ -12240,23 +12253,23 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, + "prompt_cache_min_tokens": 4096, "supports_assistant_prefill": true, + "supports_computer_use": true, "supports_function_calling": true, "supports_native_structured_output": true, - "supports_computer_use": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "claude-haiku-4-5": { - "deprecation_date": "2026-10-15", "cache_creation_input_token_cost": 1.25e-06, "cache_creation_input_token_cost_above_1hr": 2e-06, "cache_read_input_token_cost": 1e-07, + "deprecation_date": "2026-10-15", "input_cost_per_token": 1e-06, "litellm_provider": "anthropic", "max_input_tokens": 200000, @@ -12264,17 +12277,17 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, + "prompt_cache_min_tokens": 4096, "supports_assistant_prefill": true, + "supports_computer_use": true, "supports_function_calling": true, "supports_native_structured_output": true, - "supports_computer_use": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "claude-3-7-sonnet-20250219": { "cache_creation_input_token_cost": 3.75e-06, @@ -12353,6 +12366,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12366,8 +12380,7 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "claude-4-sonnet-20250514": { "cache_creation_input_token_cost": 3.75e-06, @@ -12384,6 +12397,7 @@ "mode": "chat", "output_cost_per_token": 1.5e-05, "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12398,26 +12412,26 @@ "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "prompt_cache_min_tokens": 1024 + "supports_web_search": true }, "claude-sonnet-4-5": { - "deprecation_date": "2026-09-29", "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.2e-05, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "deprecation_date": "2026-09-29", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "litellm_provider": "anthropic", "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12426,32 +12440,32 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "claude-sonnet-4-5-20250929": { - "deprecation_date": "2026-09-29", "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.2e-05, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "deprecation_date": "2026-09-29", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "litellm_provider": "anthropic", "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12460,21 +12474,20 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "prompt_cache_min_tokens": 1024 + "supports_web_search": true }, "claude-sonnet-5": { - "deprecation_date": "2027-06-30", "cache_creation_input_token_cost": 2.5e-06, "cache_creation_input_token_cost_above_1hr": 4e-06, "cache_read_input_token_cost": 2e-07, + "deprecation_date": "2027-06-30", "input_cost_per_token": 2e-06, "litellm_provider": "anthropic", "max_input_tokens": 1000000, @@ -12482,17 +12495,23 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1e-05, + "prompt_cache_min_tokens": 1024, + "provider_specific_entry": { + "us": 1.1 + }, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_native_structured_output": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -12500,19 +12519,13 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "provider_specific_entry": { - "us": 1.1 - }, - "supports_output_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "claude-sonnet-4-6": { - "deprecation_date": "2027-02-17", "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, "cache_read_input_token_cost": 3e-07, + "deprecation_date": "2027-02-17", "input_cost_per_token": 3e-06, "litellm_provider": "anthropic", "max_input_tokens": 1000000, @@ -12520,6 +12533,7 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12529,52 +12543,53 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.2e-05, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "input_cost_per_token_batches": 1.5e-06, "litellm_provider": "bedrock", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "output_cost_per_token_batches": 7.5e-06, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024, - "input_cost_per_token_batches": 1.5e-06, - "output_cost_per_token_batches": 7.5e-06 + "supports_tool_search": true, + "supports_vision": true }, "claude-opus-4-1": { "cache_creation_input_token_cost": 1.875e-05, "cache_creation_input_token_cost_above_1hr": 3e-05, "cache_read_input_token_cost": 1.5e-06, + "deprecation_date": "2026-08-05", "input_cost_per_token": 1.5e-05, "litellm_provider": "anthropic", "max_input_tokens": 200000, @@ -12582,6 +12597,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12590,28 +12606,27 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024, - "deprecation_date": "2026-08-05" + "supports_vision": true }, "claude-opus-4-1-20250805": { "cache_creation_input_token_cost": 1.875e-05, "cache_creation_input_token_cost_above_1hr": 3e-05, "cache_read_input_token_cost": 1.5e-06, - "input_cost_per_token": 1.5e-05, "deprecation_date": "2026-08-05", + "input_cost_per_token": 1.5e-05, "litellm_provider": "anthropic", "max_input_tokens": 200000, "max_output_tokens": 32000, "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12620,27 +12635,27 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "claude-opus-4-20250514": { "cache_creation_input_token_cost": 1.875e-05, "cache_creation_input_token_cost_above_1hr": 3e-05, "cache_read_input_token_cost": 1.5e-06, - "input_cost_per_token": 1.5e-05, "deprecation_date": "2026-06-15", + "input_cost_per_token": 1.5e-05, "litellm_provider": "anthropic", "max_input_tokens": 200000, "max_output_tokens": 32000, "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12654,14 +12669,13 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "claude-opus-4-5-20251101": { - "deprecation_date": "2026-11-24", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2026-11-24", "input_cost_per_token": 5e-06, "litellm_provider": "anthropic", "max_input_tokens": 200000, @@ -12669,6 +12683,7 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12677,21 +12692,20 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "claude-opus-4-5": { - "deprecation_date": "2026-11-24", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2026-11-24", "input_cost_per_token": 5e-06, "litellm_provider": "anthropic", "max_input_tokens": 200000, @@ -12699,6 +12713,7 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12707,21 +12722,20 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "claude-opus-4-6": { - "deprecation_date": "2027-02-05", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-02-05", "input_cost_per_token": 5e-06, "litellm_provider": "anthropic", "max_input_tokens": 1000000, @@ -12729,6 +12743,11 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, + "provider_specific_entry": { + "us": 1.1, + "fast": 6.0 + }, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12738,27 +12757,23 @@ "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, - "supports_tool_choice": true, - "supports_vision": true, - "provider_specific_entry": { - "us": 1.1, - "fast": 6.0 - }, - "supports_output_config": true, - "supports_max_reasoning_effort": true, "supports_speed": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_choice": true, + "supports_vision": true }, "claude-opus-4-6-20260205": { - "deprecation_date": "2027-02-05", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-02-05", "input_cost_per_token": 5e-06, "litellm_provider": "anthropic", "max_input_tokens": 1000000, @@ -12766,6 +12781,11 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, + "provider_specific_entry": { + "us": 1.1, + "fast": 6.0 + }, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12775,27 +12795,23 @@ "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, - "supports_tool_choice": true, - "supports_vision": true, - "provider_specific_entry": { - "us": 1.1, - "fast": 6.0 - }, - "supports_max_reasoning_effort": true, - "supports_output_config": true, "supports_speed": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_choice": true, + "supports_vision": true }, "claude-opus-4-7": { - "deprecation_date": "2027-04-16", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-04-16", "input_cost_per_token": 5e-06, "litellm_provider": "anthropic", "max_input_tokens": 1000000, @@ -12803,6 +12819,11 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 2048, + "provider_specific_entry": { + "us": 1.1, + "fast": 6.0 + }, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12812,29 +12833,24 @@ "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, "supports_sampling_params": false, + "supports_speed": true, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "provider_specific_entry": { - "us": 1.1, - "fast": 6.0 - }, - "supports_output_config": true, - "supports_speed": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "claude-opus-4-7-20260416": { - "deprecation_date": "2027-04-16", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-04-16", "input_cost_per_token": 5e-06, "litellm_provider": "anthropic", "max_input_tokens": 1000000, @@ -12842,6 +12858,11 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 2048, + "provider_specific_entry": { + "us": 1.1, + "fast": 6.0 + }, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12851,29 +12872,24 @@ "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, "supports_sampling_params": false, + "supports_speed": true, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "provider_specific_entry": { - "us": 1.1, - "fast": 6.0 - }, - "supports_output_config": true, - "supports_speed": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "claude-fable-5": { - "deprecation_date": "2027-06-09", "cache_creation_input_token_cost": 1.25e-05, "cache_creation_input_token_cost_above_1hr": 2e-05, "cache_read_input_token_cost": 1e-06, + "deprecation_date": "2027-06-09", "input_cost_per_token": 1e-05, "litellm_provider": "anthropic", "max_input_tokens": 1000000, @@ -12881,38 +12897,38 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 5e-05, - "search_context_cost_per_query": { - "search_context_size_high": 0.01, - "search_context_size_low": 0.01, - "search_context_size_medium": 0.01 - }, - "supports_adaptive_thinking": true, - "thinking_always_on": true, - "supports_mid_conversation_system": true, - "supports_assistant_prefill": false, - "supports_computer_use": true, - "supports_function_calling": true, - "supports_pdf_input": true, - "supports_prompt_caching": true, - "supports_reasoning": true, - "supports_response_schema": true, - "supports_sampling_params": false, - "supports_tool_choice": true, - "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, + "prompt_cache_min_tokens": 512, "provider_specific_entry": { "us": 1.1 }, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, + "supports_adaptive_thinking": true, + "supports_assistant_prefill": false, + "supports_computer_use": true, + "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, "supports_output_config": true, - "prompt_cache_min_tokens": 512, - "supports_native_structured_output": true + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_sampling_params": false, + "supports_tool_choice": true, + "supports_vision": true, + "supports_xhigh_reasoning_effort": true, + "thinking_always_on": true }, "claude-opus-5": { - "deprecation_date": "2027-07-24", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-07-24", "input_cost_per_token": 5e-06, "litellm_provider": "anthropic", "max_input_tokens": 1000000, @@ -12920,39 +12936,39 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 512, + "provider_specific_entry": { + "us": 1.1, + "fast": 2.0 + }, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, "supports_sampling_params": false, + "supports_speed": true, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "provider_specific_entry": { - "us": 1.1, - "fast": 2.0 - }, - "supports_output_config": true, - "supports_speed": true, - "prompt_cache_min_tokens": 512 + "supports_xhigh_reasoning_effort": true }, "claude-opus-4-8": { - "deprecation_date": "2027-05-28", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-05-28", "input_cost_per_token": 5e-06, "litellm_provider": "anthropic", "max_input_tokens": 1000000, @@ -12960,50 +12976,51 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 1024, + "provider_specific_entry": { + "us": 1.1, + "fast": 2.0 + }, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, "supports_sampling_params": false, + "supports_speed": true, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "provider_specific_entry": { - "us": 1.1, - "fast": 2.0 - }, - "supports_output_config": true, - "supports_speed": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "claude-sonnet-4-20250514": { - "deprecation_date": "2026-06-15", "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "deprecation_date": "2026-06-15", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "litellm_provider": "anthropic", "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -13017,8 +13034,7 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "cloudflare/@cf/meta/llama-2-7b-chat-fp16": { "input_cost_per_token": 1.923e-06, @@ -13463,14 +13479,14 @@ "output_cost_per_token": 0.0 }, "command": { + "deprecation_date": "2025-09-15", "input_cost_per_token": 1e-06, "litellm_provider": "cohere", "max_input_tokens": 4096, "max_output_tokens": 4096, "max_tokens": 4096, "mode": "completion", - "output_cost_per_token": 2e-06, - "deprecation_date": "2025-09-15" + "output_cost_per_token": 2e-06 }, "command-a-03-2025": { "input_cost_per_token": 2.5e-06, @@ -13484,6 +13500,7 @@ "supports_tool_choice": true }, "command-light": { + "deprecation_date": "2025-09-15", "input_cost_per_token": 3e-07, "litellm_provider": "cohere_chat", "max_input_tokens": 4096, @@ -13491,8 +13508,7 @@ "max_tokens": 4096, "mode": "chat", "output_cost_per_token": 6e-07, - "supports_tool_choice": true, - "deprecation_date": "2025-09-15" + "supports_tool_choice": true }, "command-nightly": { "input_cost_per_token": 1e-06, @@ -13504,6 +13520,7 @@ "output_cost_per_token": 2e-06 }, "command-r": { + "deprecation_date": "2025-09-15", "input_cost_per_token": 1.5e-07, "litellm_provider": "cohere_chat", "max_input_tokens": 128000, @@ -13512,8 +13529,7 @@ "mode": "chat", "output_cost_per_token": 6e-07, "supports_function_calling": true, - "supports_tool_choice": true, - "deprecation_date": "2025-09-15" + "supports_tool_choice": true }, "command-r-08-2024": { "input_cost_per_token": 1.5e-07, @@ -13527,6 +13543,7 @@ "supports_tool_choice": true }, "command-r-plus": { + "deprecation_date": "2025-09-15", "input_cost_per_token": 2.5e-06, "litellm_provider": "cohere_chat", "max_input_tokens": 128000, @@ -13535,8 +13552,7 @@ "mode": "chat", "output_cost_per_token": 1e-05, "supports_function_calling": true, - "supports_tool_choice": true, - "deprecation_date": "2025-09-15" + "supports_tool_choice": true }, "command-r-plus-08-2024": { "input_cost_per_token": 2.5e-06, @@ -14674,8 +14690,8 @@ "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "databricks/databricks-claude-opus-4": { "cache_creation_input_token_cost": 1.874999e-05, @@ -14719,8 +14735,8 @@ "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "databricks/databricks-claude-opus-4-5": { "cache_creation_input_token_cost": 6.25002e-06, @@ -14740,11 +14756,11 @@ "source": "https://www.databricks.com/product/pricing/proprietary-foundation-model-serving", "supports_assistant_prefill": true, "supports_function_calling": true, + "supports_output_config": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "supports_output_config": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "databricks/databricks-claude-opus-4-6": { "cache_creation_input_token_cost": 6.25002e-06, @@ -14764,6 +14780,7 @@ "source": "https://www.databricks.com/product/pricing/proprietary-foundation-model-serving", "supports_assistant_prefill": true, "supports_function_calling": true, + "supports_legacy_thinking": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_tool_choice": true @@ -14868,8 +14885,8 @@ "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "databricks/databricks-claude-sonnet-4-1": { "cache_creation_input_token_cost": 3.74997e-06, @@ -14913,8 +14930,8 @@ "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "databricks/databricks-claude-sonnet-4-6": { "cache_creation_input_token_cost": 3.74997e-06, @@ -14934,6 +14951,7 @@ "source": "https://www.databricks.com/product/pricing/proprietary-foundation-model-serving", "supports_assistant_prefill": true, "supports_function_calling": true, + "supports_legacy_thinking": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_tool_choice": true @@ -14983,8 +15001,8 @@ "source": "https://www.databricks.com/product/pricing/proprietary-foundation-model-serving", "supports_function_calling": true, "supports_prompt_caching": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "databricks/databricks-gemini-2-5-pro": { "cache_creation_input_token_cost": 1.24999e-06, @@ -15004,8 +15022,8 @@ "source": "https://www.databricks.com/product/pricing/proprietary-foundation-model-serving", "supports_function_calling": true, "supports_prompt_caching": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "databricks/databricks-gemini-3-1-flash-lite": { "cache_creation_input_token_cost": 3.1248e-07, @@ -15412,8 +15430,8 @@ "output_cost_per_token": 1.5000300000000002e-06, "output_dbu_cost_per_token": 2.1429e-05, "source": "https://www.databricks.com/product/pricing/foundation-model-serving", - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "databricks/databricks-meta-llama-3-1-405b-instruct": { "cache_creation_input_token_cost": 5.00003e-06, @@ -15467,8 +15485,8 @@ "output_cost_per_token": 1.5000300000000002e-06, "output_dbu_cost_per_token": 2.1429e-05, "source": "https://www.databricks.com/product/pricing/foundation-model-serving", - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "databricks/databricks-meta-llama-3-70b-instruct": { "cache_creation_input_token_cost": 1.00002e-06, @@ -16056,750 +16074,750 @@ ] }, "deepinfra/Gryphe/MythoMax-L2-13b": { - "max_tokens": 4096, + "input_cost_per_token": 8e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 8e-08, - "output_cost_per_token": 9e-08, - "litellm_provider": "deepinfra", + "max_tokens": 4096, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 9e-08, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/NousResearch/Hermes-3-Llama-3.1-405B": { - "max_tokens": 131072, + "input_cost_per_token": 1e-06, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-06, - "output_cost_per_token": 1e-06, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/NousResearch/Hermes-3-Llama-3.1-70B": { - "max_tokens": 131072, + "input_cost_per_token": 3e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 3e-07, - "output_cost_per_token": 3e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 3e-07, "supports_tool_choice": false }, "deepinfra/Qwen/QwQ-32B": { - "max_tokens": 131072, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 4e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Qwen/Qwen2.5-72B-Instruct": { - "max_tokens": 32768, + "input_cost_per_token": 1.2e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1.2e-07, - "output_cost_per_token": 3.9e-07, - "litellm_provider": "deepinfra", + "max_tokens": 32768, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 3.9e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Qwen/Qwen2.5-7B-Instruct": { - "max_tokens": 32768, + "input_cost_per_token": 4e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 4e-08, - "output_cost_per_token": 1e-07, - "litellm_provider": "deepinfra", + "max_tokens": 32768, "mode": "chat", + "output_cost_per_token": 1e-07, "supports_tool_choice": false }, "deepinfra/Qwen/Qwen2.5-VL-32B-Instruct": { - "max_tokens": 128000, + "input_cost_per_token": 2e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "deepinfra", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 6e-07, + "supports_function_calling": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_function_calling": true + "supports_vision": true }, "deepinfra/Qwen/Qwen3-14B": { - "max_tokens": 40960, + "input_cost_per_token": 6e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 6e-08, - "output_cost_per_token": 2.4e-07, - "litellm_provider": "deepinfra", + "max_tokens": 40960, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2.4e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Qwen/Qwen3-235B-A22B": { - "max_tokens": 40960, + "input_cost_per_token": 1.8e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 1.8e-07, - "output_cost_per_token": 5.4e-07, - "litellm_provider": "deepinfra", + "max_tokens": 40960, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 5.4e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Qwen/Qwen3-235B-A22B-Instruct-2507": { - "max_tokens": 262144, + "input_cost_per_token": 9e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 9e-08, - "output_cost_per_token": 6e-07, - "litellm_provider": "deepinfra", + "max_tokens": 262144, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 6e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Qwen/Qwen3-235B-A22B-Thinking-2507": { - "max_tokens": 262144, + "input_cost_per_token": 3e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 3e-07, - "output_cost_per_token": 2.9e-06, - "litellm_provider": "deepinfra", + "max_tokens": 262144, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2.9e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Qwen/Qwen3-30B-A3B": { - "max_tokens": 40960, + "input_cost_per_token": 8e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 8e-08, - "output_cost_per_token": 2.9e-07, - "litellm_provider": "deepinfra", + "max_tokens": 40960, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2.9e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Qwen/Qwen3-32B": { - "max_tokens": 40960, + "input_cost_per_token": 1e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 2.8e-07, - "litellm_provider": "deepinfra", + "max_tokens": 40960, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2.8e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Qwen/Qwen3-Coder-480B-A35B-Instruct": { - "max_tokens": 262144, + "input_cost_per_token": 4e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 4e-07, - "output_cost_per_token": 1.6e-06, - "litellm_provider": "deepinfra", + "max_tokens": 262144, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1.6e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Qwen/Qwen3-Coder-480B-A35B-Instruct-Turbo": { - "max_tokens": 262144, + "input_cost_per_token": 2.9e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 2.9e-07, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "deepinfra", + "max_tokens": 262144, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1.2e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Qwen/Qwen3-Next-80B-A3B-Instruct": { - "max_tokens": 262144, + "input_cost_per_token": 1.4e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1.4e-07, - "output_cost_per_token": 1.4e-06, - "litellm_provider": "deepinfra", + "max_tokens": 262144, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1.4e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Qwen/Qwen3-Next-80B-A3B-Thinking": { - "max_tokens": 262144, + "input_cost_per_token": 1.4e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1.4e-07, - "output_cost_per_token": 1.4e-06, - "litellm_provider": "deepinfra", + "max_tokens": 262144, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1.4e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Sao10K/L3-8B-Lunaris-v1-Turbo": { - "max_tokens": 8192, + "input_cost_per_token": 4e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 4e-08, - "output_cost_per_token": 5e-08, - "litellm_provider": "deepinfra", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 5e-08, "supports_tool_choice": false }, "deepinfra/Sao10K/L3.1-70B-Euryale-v2.2": { - "max_tokens": 131072, + "input_cost_per_token": 6.5e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 6.5e-07, - "output_cost_per_token": 7.5e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 7.5e-07, "supports_tool_choice": false }, "deepinfra/Sao10K/L3.3-70B-Euryale-v2.3": { - "max_tokens": 131072, + "input_cost_per_token": 6.5e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 6.5e-07, - "output_cost_per_token": 7.5e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 7.5e-07, "supports_tool_choice": false }, "deepinfra/allenai/olmOCR-7B-0725-FP8": { - "max_tokens": 16384, + "input_cost_per_token": 2.7e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 2.7e-07, - "output_cost_per_token": 1.5e-06, - "litellm_provider": "deepinfra", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 1.5e-06, "supports_tool_choice": false }, "deepinfra/anthropic/claude-3-7-sonnet-latest": { - "max_tokens": 200000, + "cache_read_input_token_cost": 3.3e-07, + "input_cost_per_token": 3.3e-06, + "litellm_provider": "deepinfra", "max_input_tokens": 200000, "max_output_tokens": 200000, - "input_cost_per_token": 3.3e-06, - "output_cost_per_token": 1.65e-05, - "cache_read_input_token_cost": 3.3e-07, - "litellm_provider": "deepinfra", + "max_tokens": 200000, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1.65e-05, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/anthropic/claude-4-opus": { - "max_tokens": 200000, + "input_cost_per_token": 1.65e-05, + "litellm_provider": "deepinfra", "max_input_tokens": 200000, "max_output_tokens": 200000, - "input_cost_per_token": 1.65e-05, - "output_cost_per_token": 8.25e-05, - "litellm_provider": "deepinfra", + "max_tokens": 200000, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 8.25e-05, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/anthropic/claude-4-sonnet": { - "max_tokens": 200000, + "input_cost_per_token": 3.3e-06, + "litellm_provider": "deepinfra", "max_input_tokens": 200000, "max_output_tokens": 200000, - "input_cost_per_token": 3.3e-06, - "output_cost_per_token": 1.65e-05, - "litellm_provider": "deepinfra", + "max_tokens": 200000, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1.65e-05, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/deepseek-ai/DeepSeek-R1": { - "max_tokens": 163840, + "input_cost_per_token": 7e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 7e-07, - "output_cost_per_token": 2.4e-06, - "litellm_provider": "deepinfra", + "max_tokens": 163840, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2.4e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/deepseek-ai/DeepSeek-R1-0528": { - "max_tokens": 163840, + "cache_read_input_token_cost": 4e-07, + "input_cost_per_token": 5e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 2.15e-06, - "cache_read_input_token_cost": 4e-07, - "litellm_provider": "deepinfra", + "max_tokens": 163840, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2.15e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/deepseek-ai/DeepSeek-R1-0528-Turbo": { - "max_tokens": 32768, + "input_cost_per_token": 1e-06, + "litellm_provider": "deepinfra", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-06, - "output_cost_per_token": 3e-06, - "litellm_provider": "deepinfra", + "max_tokens": 32768, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 3e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/deepseek-ai/DeepSeek-R1-Distill-Llama-70B": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 6e-07, "supports_tool_choice": false }, "deepinfra/deepseek-ai/DeepSeek-R1-Distill-Qwen-32B": { - "max_tokens": 131072, + "input_cost_per_token": 2.7e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2.7e-07, - "output_cost_per_token": 2.7e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2.7e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/deepseek-ai/DeepSeek-R1-Turbo": { - "max_tokens": 40960, + "input_cost_per_token": 1e-06, + "litellm_provider": "deepinfra", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 1e-06, - "output_cost_per_token": 3e-06, - "litellm_provider": "deepinfra", + "max_tokens": 40960, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 3e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/deepseek-ai/DeepSeek-V3": { - "max_tokens": 163840, + "input_cost_per_token": 3.8e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 3.8e-07, - "output_cost_per_token": 8.9e-07, - "litellm_provider": "deepinfra", + "max_tokens": 163840, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 8.9e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/deepseek-ai/DeepSeek-V3-0324": { - "max_tokens": 163840, + "input_cost_per_token": 2.5e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 2.5e-07, - "output_cost_per_token": 8.8e-07, - "litellm_provider": "deepinfra", + "max_tokens": 163840, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 8.8e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/deepseek-ai/DeepSeek-V3.1": { - "max_tokens": 163840, + "cache_read_input_token_cost": 2.16e-07, + "input_cost_per_token": 2.7e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 2.7e-07, - "output_cost_per_token": 1e-06, - "cache_read_input_token_cost": 2.16e-07, - "litellm_provider": "deepinfra", + "max_tokens": 163840, "mode": "chat", - "supports_tool_choice": true, + "output_cost_per_token": 1e-06, + "supports_function_calling": true, "supports_reasoning": true, - "supports_function_calling": true + "supports_tool_choice": true }, "deepinfra/deepseek-ai/DeepSeek-V3.1-Terminus": { - "max_tokens": 163840, + "cache_read_input_token_cost": 2.16e-07, + "input_cost_per_token": 2.7e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 2.7e-07, - "output_cost_per_token": 1e-06, - "cache_read_input_token_cost": 2.16e-07, - "litellm_provider": "deepinfra", + "max_tokens": 163840, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/google/gemini-2.0-flash-001": { "deprecation_date": "2026-06-01", - "max_tokens": 1000000, + "input_cost_per_token": 1e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 1000000, "max_output_tokens": 1000000, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "deepinfra", + "max_tokens": 1000000, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 4e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/google/gemini-2.5-flash": { - "max_tokens": 1000000, + "input_cost_per_token": 3e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 1000000, "max_output_tokens": 1000000, - "input_cost_per_token": 3e-07, - "output_cost_per_token": 2.5e-06, - "litellm_provider": "deepinfra", + "max_tokens": 1000000, "mode": "chat", - "supports_tool_choice": true, + "output_cost_per_token": 2.5e-06, "supports_function_calling": true, - "supports_image_size": false + "supports_image_size": false, + "supports_tool_choice": true }, "deepinfra/google/gemini-2.5-pro": { - "max_tokens": 1000000, + "input_cost_per_token": 1.25e-06, + "litellm_provider": "deepinfra", "max_input_tokens": 1000000, "max_output_tokens": 1000000, - "input_cost_per_token": 1.25e-06, - "output_cost_per_token": 1e-05, - "litellm_provider": "deepinfra", + "max_tokens": 1000000, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1e-05, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/google/gemma-3-12b-it": { - "max_tokens": 131072, + "input_cost_per_token": 5e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 5e-08, - "output_cost_per_token": 1e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/google/gemma-3-27b-it": { - "max_tokens": 131072, + "input_cost_per_token": 9e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-08, - "output_cost_per_token": 1.6e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1.6e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/google/gemma-3-4b-it": { - "max_tokens": 131072, + "input_cost_per_token": 4e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 4e-08, - "output_cost_per_token": 8e-08, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 8e-08, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/meta-llama/Llama-3.2-11B-Vision-Instruct": { - "max_tokens": 131072, + "input_cost_per_token": 4.9e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 4.9e-08, - "output_cost_per_token": 4.9e-08, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 4.9e-08, "supports_tool_choice": false }, "deepinfra/meta-llama/Llama-3.2-3B-Instruct": { - "max_tokens": 131072, + "input_cost_per_token": 2e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-08, - "output_cost_per_token": 2e-08, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2e-08, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/meta-llama/Llama-3.3-70B-Instruct": { - "max_tokens": 131072, + "input_cost_per_token": 2.3e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2.3e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 4e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/meta-llama/Llama-3.3-70B-Instruct-Turbo": { - "max_tokens": 131072, + "input_cost_per_token": 1.3e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1.3e-07, - "output_cost_per_token": 3.9e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 3.9e-07, "supports_function_calling": true, "supports_tool_choice": true }, "deepinfra/meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8": { - "max_tokens": 1048576, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 1048576, "max_output_tokens": 1048576, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "deepinfra", + "max_tokens": 1048576, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 6e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/meta-llama/Llama-4-Scout-17B-16E-Instruct": { - "max_tokens": 327680, + "input_cost_per_token": 8e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 327680, "max_output_tokens": 327680, - "input_cost_per_token": 8e-08, - "output_cost_per_token": 3e-07, - "litellm_provider": "deepinfra", + "max_tokens": 327680, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 3e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/meta-llama/Llama-Guard-3-8B": { - "max_tokens": 131072, + "input_cost_per_token": 5.5e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 5.5e-08, - "output_cost_per_token": 5.5e-08, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 5.5e-08, "supports_tool_choice": false }, "deepinfra/meta-llama/Llama-Guard-4-12B": { - "max_tokens": 163840, + "input_cost_per_token": 1.8e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 1.8e-07, - "output_cost_per_token": 1.8e-07, - "litellm_provider": "deepinfra", + "max_tokens": 163840, "mode": "chat", + "output_cost_per_token": 1.8e-07, "supports_tool_choice": false }, "deepinfra/meta-llama/Meta-Llama-3-8B-Instruct": { - "max_tokens": 8192, + "input_cost_per_token": 3e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 3e-08, - "output_cost_per_token": 6e-08, - "litellm_provider": "deepinfra", + "max_tokens": 8192, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 6e-08, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/meta-llama/Meta-Llama-3.1-70B-Instruct": { - "max_tokens": 131072, + "input_cost_per_token": 4e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 4e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 4e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo": { - "max_tokens": 131072, + "input_cost_per_token": 1e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 2.8e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2.8e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/meta-llama/Meta-Llama-3.1-8B-Instruct": { - "max_tokens": 131072, + "input_cost_per_token": 3e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 3e-08, - "output_cost_per_token": 5e-08, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 5e-08, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo": { - "max_tokens": 131072, + "input_cost_per_token": 2e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-08, - "output_cost_per_token": 3e-08, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 3e-08, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/microsoft/WizardLM-2-8x22B": { - "max_tokens": 65536, + "input_cost_per_token": 4.8e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 65536, "max_output_tokens": 65536, - "input_cost_per_token": 4.8e-07, - "output_cost_per_token": 4.8e-07, - "litellm_provider": "deepinfra", + "max_tokens": 65536, "mode": "chat", + "output_cost_per_token": 4.8e-07, "supports_tool_choice": false }, "deepinfra/microsoft/phi-4": { - "max_tokens": 16384, + "input_cost_per_token": 7e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 7e-08, - "output_cost_per_token": 1.4e-07, - "litellm_provider": "deepinfra", + "max_tokens": 16384, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1.4e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/mistralai/Mistral-Nemo-Instruct-2407": { - "max_tokens": 131072, + "input_cost_per_token": 2e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-08, - "output_cost_per_token": 4e-08, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 4e-08, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/mistralai/Mistral-Small-24B-Instruct-2501": { - "max_tokens": 32768, + "input_cost_per_token": 5e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 5e-08, - "output_cost_per_token": 8e-08, - "litellm_provider": "deepinfra", + "max_tokens": 32768, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 8e-08, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/mistralai/Mistral-Small-3.2-24B-Instruct-2506": { - "max_tokens": 128000, + "input_cost_per_token": 7.5e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 7.5e-08, - "output_cost_per_token": 2e-07, - "litellm_provider": "deepinfra", + "max_tokens": 128000, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/mistralai/Mixtral-8x7B-Instruct-v0.1": { - "max_tokens": 32768, + "input_cost_per_token": 4e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 4e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "deepinfra", + "max_tokens": 32768, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 4e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/moonshotai/Kimi-K2-Instruct": { - "max_tokens": 131072, + "input_cost_per_token": 5e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 2e-06, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/moonshotai/Kimi-K2-Instruct-0905": { - "max_tokens": 262144, + "cache_read_input_token_cost": 4e-07, + "input_cost_per_token": 5e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 2e-06, - "cache_read_input_token_cost": 4e-07, - "litellm_provider": "deepinfra", + "max_tokens": 262144, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/nvidia/Llama-3.1-Nemotron-70B-Instruct": { - "max_tokens": 131072, + "input_cost_per_token": 6e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 6e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 6e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/nvidia/Llama-3.3-Nemotron-Super-49B-v1.5": { - "max_tokens": 131072, + "input_cost_per_token": 1e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 4e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/nvidia/NVIDIA-Nemotron-3.5-Lightning": { - "max_input_tokens": 262144, "input_cost_per_token": 5e-08, - "output_cost_per_token": 2e-07, "litellm_provider": "deepinfra", + "max_input_tokens": 262144, "mode": "chat", + "output_cost_per_token": 2e-07, "source": "https://deepinfra.com/nvidia/NVIDIA-Nemotron-3.5-Lightning", - "supports_tool_choice": true, "supports_function_calling": true, - "supports_reasoning": true + "supports_reasoning": true, + "supports_tool_choice": true }, "deepinfra/nvidia/NVIDIA-Nemotron-Nano-9B-v2": { - "max_tokens": 131072, + "input_cost_per_token": 4e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 4e-08, - "output_cost_per_token": 1.6e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1.6e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/openai/gpt-oss-120b": { - "max_tokens": 131072, + "input_cost_per_token": 5e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 5e-08, - "output_cost_per_token": 4.5e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 4.5e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/openai/gpt-oss-20b": { - "max_tokens": 131072, + "input_cost_per_token": 4e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 4e-08, - "output_cost_per_token": 1.5e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1.5e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/zai-org/GLM-4.5": { - "max_tokens": 131072, + "input_cost_per_token": 4e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 4e-07, - "output_cost_per_token": 1.6e-06, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1.6e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepseek/deepseek-chat": { "cache_creation_input_token_cost": 0.0, @@ -16918,9 +16936,9 @@ "mode": "chat", "output_cost_per_token": 1.68e-06, "supports_function_calling": true, + "supports_native_structured_output": true, "supports_reasoning": true, - "supports_tool_choice": true, - "supports_native_structured_output": true + "supports_tool_choice": true }, "deepseek.v3.2": { "input_cost_per_token": 6.2e-07, @@ -16930,11 +16948,11 @@ "max_tokens": 163840, "mode": "chat", "output_cost_per_token": 1.85e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_reasoning": true, + "supports_tool_choice": true }, "dolphin": { "input_cost_per_token": 5e-07, @@ -17069,6 +17087,9 @@ }, "firecrawl/search": { "litellm_provider": "firecrawl", + "metadata": { + "notes": "Firecrawl search pricing: $83 for 100,000 credits, 2 credits per 10 results. Cost = ceiling(limit/10) * 2 * $0.00083" + }, "mode": "search", "tiered_pricing": [ { @@ -17141,10 +17162,7 @@ 100 ] } - ], - "metadata": { - "notes": "Firecrawl search pricing: $83 for 100,000 credits, 2 credits per 10 results. Cost = ceiling(limit/10) * 2 * $0.00083" - } + ] }, "perplexity/search": { "input_cost_per_query": 0.005, @@ -17152,60 +17170,60 @@ "mode": "search" }, "searxng/search": { - "litellm_provider": "searxng", - "mode": "search", "input_cost_per_query": 0.0, + "litellm_provider": "searxng", "metadata": { "notes": "SearXNG is an open-source metasearch engine. Free to use when self-hosted or using public instances." - } + }, + "mode": "search" }, "serper/search": { "input_cost_per_query": 0.001, "litellm_provider": "serper", - "mode": "search", "metadata": { "notes": "Serper Google Search API. Pricing: $1.00/1k queries (Starter), $0.75/1k (Standard), $0.50/1k (Scale), $0.30/1k (Ultimate)." - } + }, + "mode": "search" }, "apiserpent/search": { "input_cost_per_query": 0.0006, "litellm_provider": "apiserpent", - "mode": "search", "metadata": { "notes": "APISerpent quick search (/api/search/quick), multi-engine (Google, Bing, Yahoo, DuckDuckGo). Pricing: $0.60/1k searches." - } + }, + "mode": "search" }, "apiserpent/deep_search": { "input_cost_per_query": 0.0006, "litellm_provider": "apiserpent", - "mode": "search", "metadata": { "notes": "APISerpent deep search (/api/search), multi-engine (Google, Bing, Yahoo, DuckDuckGo). Pricing: $0.60/1k searches." - } + }, + "mode": "search" }, "agentcore/search": { "input_cost_per_query": 0.0, "litellm_provider": "agentcore", - "mode": "search", "metadata": { "notes": "Web Search on Amazon Bedrock AgentCore, billed by AWS on the gateway" - } + }, + "mode": "search" }, "tinyfish/search": { "input_cost_per_query": 0.0, "litellm_provider": "tinyfish", - "mode": "search", "metadata": { "notes": "TinyFish Search API" - } + }, + "mode": "search" }, "nimble/search": { "input_cost_per_query": 0.005, "litellm_provider": "nimble", - "mode": "search", "metadata": { "notes": "Nimble Search API pay-as-you-go list price: $5 per 1,000 searches, up to 100 results per search. Volume plans price differently." - } + }, + "mode": "search" }, "elevenlabs/scribe_v1": { "input_cost_per_second": 6.11e-05, @@ -17372,6 +17390,8 @@ "supports_vision": true }, "eu.anthropic.claude-3-5-haiku-20241022-v1:0": { + "cache_creation_input_token_cost": 3.125e-07, + "cache_read_input_token_cost": 2.5e-08, "input_cost_per_token": 2.5e-07, "litellm_provider": "bedrock", "max_input_tokens": 200000, @@ -17379,46 +17399,46 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.25e-06, + "prompt_cache_min_tokens": 2048, "supports_assistant_prefill": true, "supports_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_response_schema": true, - "supports_tool_choice": true, - "cache_read_input_token_cost": 2.5e-08, - "cache_creation_input_token_cost": 3.125e-07, - "prompt_cache_min_tokens": 2048 + "supports_tool_choice": true }, "eu.anthropic.claude-haiku-4-5-20251001-v1:0": { "cache_creation_input_token_cost": 1.375e-06, "cache_creation_input_token_cost_above_1hr": 2.2e-06, "cache_read_input_token_cost": 1.1e-07, - "input_cost_per_token": 1.1e-06, "deprecation_date": "2026-10-15", + "input_cost_per_token": 1.1e-06, + "input_cost_per_token_batches": 5.5e-07, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5.5e-06, + "output_cost_per_token_batches": 2.75e-06, + "prompt_cache_min_tokens": 4096, "source": "https://aws.amazon.com/about-aws/whats-new/2025/10/claude-4-5-haiku-anthropic-amazon-bedrock", "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096, - "input_cost_per_token_batches": 5.5e-07, - "output_cost_per_token_batches": 2.75e-06 + "supports_tool_search": true, + "supports_vision": true }, "eu.anthropic.claude-3-5-sonnet-20240620-v1:0": { + "cache_creation_input_token_cost": 3.75e-06, + "cache_read_input_token_cost": 3e-07, "input_cost_per_token": 3e-06, "litellm_provider": "bedrock", "max_input_tokens": 200000, @@ -17430,11 +17450,11 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-07, - "cache_creation_input_token_cost": 3.75e-06 + "supports_vision": true }, "eu.anthropic.claude-3-5-sonnet-20241022-v2:0": { + "cache_creation_input_token_cost": 3.75e-06, + "cache_read_input_token_cost": 3e-07, "input_cost_per_token": 3e-06, "litellm_provider": "bedrock", "max_input_tokens": 200000, @@ -17449,11 +17469,11 @@ "supports_prompt_caching": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-07, - "cache_creation_input_token_cost": 3.75e-06 + "supports_vision": true }, "eu.anthropic.claude-3-7-sonnet-20250219-v1:0": { + "cache_creation_input_token_cost": 3.75e-06, + "cache_read_input_token_cost": 3e-07, "input_cost_per_token": 3e-06, "litellm_provider": "bedrock", "max_input_tokens": 200000, @@ -17469,11 +17489,11 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-07, - "cache_creation_input_token_cost": 3.75e-06 + "supports_vision": true }, "eu.anthropic.claude-3-haiku-20240307-v1:0": { + "cache_creation_input_token_cost": 3.125e-07, + "cache_read_input_token_cost": 2.5e-08, "deprecation_date": "2026-09-10", "input_cost_per_token": 2.5e-07, "litellm_provider": "bedrock", @@ -17486,11 +17506,11 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 2.5e-08, - "cache_creation_input_token_cost": 3.125e-07 + "supports_vision": true }, "eu.anthropic.claude-3-opus-20240229-v1:0": { + "cache_creation_input_token_cost": 1.875e-05, + "cache_read_input_token_cost": 1.5e-06, "input_cost_per_token": 1.5e-05, "litellm_provider": "bedrock", "max_input_tokens": 200000, @@ -17501,11 +17521,11 @@ "supports_function_calling": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 1.5e-06, - "cache_creation_input_token_cost": 1.875e-05 + "supports_vision": true }, "eu.anthropic.claude-3-sonnet-20240229-v1:0": { + "cache_creation_input_token_cost": 3.75e-06, + "cache_read_input_token_cost": 3e-07, "deprecation_date": "2026-07-30", "input_cost_per_token": 3e-06, "litellm_provider": "bedrock", @@ -17518,9 +17538,7 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-07, - "cache_creation_input_token_cost": 3.75e-06 + "supports_vision": true }, "eu.anthropic.claude-opus-4-1-20250805-v1:0": { "cache_creation_input_token_cost": 1.875e-05, @@ -17532,6 +17550,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -17545,8 +17564,7 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "eu.anthropic.claude-opus-4-20250514-v1:0": { "cache_creation_input_token_cost": 1.875e-05, @@ -17558,6 +17576,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -17571,24 +17590,25 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "eu.anthropic.claude-sonnet-4-20250514-v1:0": { + "bedrock_converse_supports_strict_tools": false, "cache_creation_input_token_cost": 3.75e-06, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, "deprecation_date": "2026-10-14", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -17602,27 +17622,27 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "bedrock_converse_supports_strict_tools": false, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "eu.anthropic.claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 4.125e-06, "cache_creation_input_token_cost_above_1hr": 6.6e-06, + "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.32e-05, + "cache_creation_input_token_cost_above_200k_tokens": 8.25e-06, "cache_read_input_token_cost": 3.3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6.6e-07, "input_cost_per_token": 3.3e-06, "input_cost_per_token_above_200k_tokens": 6.6e-06, - "output_cost_per_token_above_200k_tokens": 2.475e-05, - "cache_creation_input_token_cost_above_200k_tokens": 8.25e-06, - "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.32e-05, - "cache_read_input_token_cost_above_200k_tokens": 6.6e-07, + "input_cost_per_token_batches": 1.65e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.65e-05, + "output_cost_per_token_above_200k_tokens": 2.475e-05, + "output_cost_per_token_batches": 8.25e-06, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -17631,17 +17651,15 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024, - "input_cost_per_token_batches": 1.65e-06, - "output_cost_per_token_batches": 8.25e-06 + "supports_tool_search": true, + "supports_vision": true }, "eu.meta.llama3-2-1b-instruct-v1:0": { "input_cost_per_token": 1.3e-07, @@ -18618,12 +18636,12 @@ }, "fireworks_ai/accounts/fireworks/models/glm-4p6": { "input_cost_per_token": 5.5e-07, - "output_cost_per_token": 2.19e-06, "litellm_provider": "fireworks_ai", "max_input_tokens": 202800, "max_output_tokens": 202800, "max_tokens": 202800, "mode": "chat", + "output_cost_per_token": 2.19e-06, "source": "https://fireworks.ai/pricing", "supports_function_calling": true, "supports_reasoning": true, @@ -19542,6 +19560,11 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 4e-07, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/pricing#2_0flash", "supported_modalities": [ "text", @@ -19563,12 +19586,7 @@ "supports_tool_choice": true, "supports_url_context": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "supports_web_search": true }, "gemini-2.0-flash-001": { "cache_read_input_token_cost": 3.75e-08, @@ -19581,6 +19599,11 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 6e-07, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_modalities": [ "text", @@ -19600,12 +19623,7 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "supports_web_search": true }, "gemini-2.0-flash-lite": { "cache_read_input_token_cost": 1.875e-08, @@ -19617,6 +19635,11 @@ "max_output_tokens": 8192, "mode": "chat", "output_cost_per_token": 3e-07, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/models#gemini-2.0-flash", "supported_modalities": [ "text", @@ -19635,12 +19658,7 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "supports_web_search": true }, "gemini-2.0-flash-lite-001": { "cache_read_input_token_cost": 1.875e-08, @@ -19652,6 +19670,11 @@ "max_output_tokens": 8192, "mode": "chat", "output_cost_per_token": 3e-07, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/models#gemini-2.0-flash", "supported_modalities": [ "text", @@ -19670,16 +19693,11 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "supports_web_search": true }, "gemini-2.5-flash": { - "deprecation_date": "2026-10-20", "cache_read_input_token_cost": 3e-08, + "deprecation_date": "2026-10-20", "input_cost_per_audio_token": 1e-06, "input_cost_per_token": 3e-07, "litellm_provider": "vertex_ai-language-models", @@ -19689,6 +19707,11 @@ "mode": "chat", "output_cost_per_reasoning_token": 2.5e-06, "output_cost_per_token": 2.5e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-preview", "supported_endpoints": [ "/v1/chat/completions", @@ -19706,6 +19729,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -19715,17 +19739,11 @@ "supports_tool_choice": true, "supports_url_context": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "supports_image_size": false + "supports_web_search": true }, "gemini-2.5-flash-image": { - "deprecation_date": "2026-10-02", "cache_read_input_token_cost": 3e-08, + "deprecation_date": "2026-10-02", "input_cost_per_audio_token": 1e-06, "input_cost_per_token": 3e-07, "litellm_provider": "vertex_ai-language-models", @@ -19756,6 +19774,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -19765,8 +19784,7 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": false, - "tpm": 8000000, - "supports_image_size": false + "tpm": 8000000 }, "gemini-3-pro-image": { "deprecation_date": "2027-05-28", @@ -19782,6 +19800,11 @@ "output_cost_per_image_token": 0.00012, "output_cost_per_token": 1.2e-05, "output_cost_per_token_batches": 6e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-3-pro-image", "supported_endpoints": [ "/v1/chat/completions", @@ -19799,16 +19822,11 @@ "supports_function_calling": false, "supports_prompt_caching": true, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_vision": true, "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, - "web_search_billing_unit": "per_query", - "supports_service_tier": true + "web_search_billing_unit": "per_query" }, "gemini-3-pro-image-preview": { "input_cost_per_image": 0.0011, @@ -19823,6 +19841,11 @@ "output_cost_per_image_token": 0.00012, "output_cost_per_token": 1.2e-05, "output_cost_per_token_batches": 6e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -19843,11 +19866,6 @@ "supports_system_messages": true, "supports_vision": true, "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini-3.1-flash-image": { @@ -19862,6 +19880,11 @@ "output_cost_per_image": 0.0672, "output_cost_per_image_token": 6e-05, "output_cost_per_token": 3e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models", "supported_endpoints": [ "/v1/chat/completions", @@ -19882,11 +19905,6 @@ "supports_system_messages": true, "supports_vision": true, "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini-3.1-flash-image-preview": { @@ -19900,6 +19918,11 @@ "output_cost_per_image": 0.0672, "output_cost_per_image_token": 6e-05, "output_cost_per_token": 3e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models", "supported_endpoints": [ "/v1/chat/completions", @@ -19920,11 +19943,6 @@ "supports_system_messages": true, "supports_vision": true, "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini-3.1-flash-lite-image": { @@ -19967,6 +19985,7 @@ }, "gemini-3.1-flash-lite-preview": { "cache_read_input_token_cost": 2.5e-08, + "deprecation_date": "2026-05-25", "input_cost_per_audio_token": 5e-07, "input_cost_per_token": 2.5e-07, "litellm_provider": "vertex_ai-language-models", @@ -19976,6 +19995,11 @@ "mode": "chat", "output_cost_per_reasoning_token": 1.5e-06, "output_cost_per_token": 1.5e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/models", "supported_endpoints": [ "/v1/chat/completions", @@ -19994,6 +20018,7 @@ "supports_audio_input": true, "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -20005,35 +20030,38 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, - "web_search_billing_unit": "per_query", - "deprecation_date": "2026-05-25" + "web_search_billing_unit": "per_query" }, "gemini-3.1-flash-lite": { - "deprecation_date": "2027-05-07", "cache_read_input_token_cost": 2.5e-08, "cache_read_input_token_cost_flex": 1.25e-08, + "cache_read_input_token_cost_per_audio_token": 5e-08, "cache_read_input_token_cost_priority": 4.5e-08, + "deprecation_date": "2027-05-07", "input_cost_per_audio_token": 5e-07, "input_cost_per_token": 2.5e-07, "input_cost_per_token_batches": 1.25e-07, "input_cost_per_token_flex": 1.25e-07, "input_cost_per_token_priority": 4.5e-07, "litellm_provider": "vertex_ai-language-models", + "max_audio_length_hours": 8.4, + "max_audio_per_prompt": 1, + "max_images_per_prompt": 3000, "max_input_tokens": 1048576, "max_output_tokens": 65536, "max_tokens": 65536, + "max_videos_per_prompt": 10, "mode": "chat", "output_cost_per_reasoning_token": 1.5e-06, "output_cost_per_token": 1.5e-06, "output_cost_per_token_batches": 7.5e-07, "output_cost_per_token_flex": 7.5e-07, "output_cost_per_token_priority": 2.7e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-3.1-flash-lite", "supported_endpoints": [ "/v1/chat/completions", @@ -20051,39 +20079,29 @@ ], "supports_audio_input": true, "supports_audio_output": false, + "supports_code_execution": true, + "supports_file_search": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_tool_choice": true, "supports_url_context": true, "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, - "web_search_billing_unit": "per_query", - "cache_read_input_token_cost_per_audio_token": 5e-08, - "max_audio_length_hours": 8.4, - "max_audio_per_prompt": 1, - "max_images_per_prompt": 3000, - "max_videos_per_prompt": 10, - "supports_code_execution": true, - "supports_file_search": true, - "supports_service_tier": true + "web_search_billing_unit": "per_query" }, "gemini-3.5-flash-lite": { - "deprecation_date": "2027-07-21", "cache_read_input_token_cost": 3e-08, "cache_read_input_token_cost_flex": 2e-08, "cache_read_input_token_cost_priority": 5e-08, + "deprecation_date": "2027-07-21", "input_cost_per_token": 3e-07, "input_cost_per_token_batches": 1.5e-07, "input_cost_per_token_flex": 1.5e-07, @@ -20098,6 +20116,11 @@ "output_cost_per_token_batches": 1.25e-06, "output_cost_per_token_flex": 1.25e-06, "output_cost_per_token_priority": 4.5e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -20116,6 +20139,7 @@ "supports_audio_input": true, "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -20127,12 +20151,6 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "deep-research-pro-preview-12-2025": { @@ -20170,8 +20188,8 @@ "supports_web_search": true }, "gemini-2.5-flash-lite": { - "deprecation_date": "2026-10-20", "cache_read_input_token_cost": 1e-08, + "deprecation_date": "2026-10-20", "input_cost_per_audio_token": 3e-07, "input_cost_per_token": 1e-07, "litellm_provider": "vertex_ai-language-models", @@ -20181,6 +20199,11 @@ "mode": "chat", "output_cost_per_reasoning_token": 4e-07, "output_cost_per_token": 4e-07, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-preview", "supported_endpoints": [ "/v1/chat/completions", @@ -20198,6 +20221,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -20207,13 +20231,7 @@ "supports_tool_choice": true, "supports_url_context": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "supports_image_size": false + "supports_web_search": true }, "gemini-2.5-flash-lite-preview-09-2025": { "cache_read_input_token_cost": 1e-08, @@ -20226,6 +20244,11 @@ "mode": "chat", "output_cost_per_reasoning_token": 4e-07, "output_cost_per_token": 4e-07, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://developers.googleblog.com/en/continuing-to-bring-you-our-latest-models-with-an-improved-gemini-2-5-flash-and-flash-lite-release/", "supported_endpoints": [ "/v1/chat/completions", @@ -20243,6 +20266,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -20252,13 +20276,7 @@ "supports_tool_choice": true, "supports_url_context": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "supports_image_size": false + "supports_web_search": true }, "gemini-2.5-flash-preview-09-2025": { "cache_read_input_token_cost": 7.5e-08, @@ -20271,6 +20289,11 @@ "mode": "chat", "output_cost_per_reasoning_token": 2.5e-06, "output_cost_per_token": 2.5e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://developers.googleblog.com/en/continuing-to-bring-you-our-latest-models-with-an-improved-gemini-2-5-flash-and-flash-lite-release/", "supported_endpoints": [ "/v1/chat/completions", @@ -20288,6 +20311,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -20297,16 +20321,11 @@ "supports_tool_choice": true, "supports_url_context": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "supports_image_size": false + "supports_web_search": true }, "gemini-live-2.5-flash-preview-native-audio-09-2025": { "cache_read_input_token_cost": 7.5e-08, + "gemini_native_audio": true, "input_cost_per_audio_token": 3e-06, "input_cost_per_token": 3e-07, "litellm_provider": "vertex_ai-language-models", @@ -20316,6 +20335,11 @@ "mode": "realtime", "output_cost_per_audio_token": 1.2e-05, "output_cost_per_token": 2e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_endpoints": [ "/vertex_ai/live" @@ -20341,16 +20365,11 @@ "supports_tool_choice": true, "supports_url_context": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "gemini_native_audio": true + "supports_web_search": true }, "gemini/gemini-live-2.5-flash-preview-native-audio-09-2025": { "cache_read_input_token_cost": 7.5e-08, + "gemini_native_audio": true, "input_cost_per_audio_token": 3e-06, "input_cost_per_token": 3e-07, "litellm_provider": "gemini", @@ -20361,6 +20380,11 @@ "output_cost_per_audio_token": 1.2e-05, "output_cost_per_token": 2e-06, "rpm": 100000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_endpoints": [ "/v1/realtime" @@ -20387,17 +20411,11 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 8000000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "gemini_native_audio": true + "tpm": 8000000 }, "gemini-2.5-flash-lite-preview-06-17": { - "deprecation_date": "2025-11-18", "cache_read_input_token_cost": 2.5e-08, + "deprecation_date": "2025-11-18", "input_cost_per_audio_token": 5e-07, "input_cost_per_token": 1e-07, "litellm_provider": "vertex_ai-language-models", @@ -20407,6 +20425,11 @@ "mode": "chat", "output_cost_per_reasoning_token": 4e-07, "output_cost_per_token": 4e-07, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-preview", "supported_endpoints": [ "/v1/chat/completions", @@ -20424,6 +20447,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -20433,19 +20457,13 @@ "supports_tool_choice": true, "supports_url_context": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "supports_image_size": false + "supports_web_search": true }, "gemini-2.5-pro": { - "deprecation_date": "2026-10-20", + "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "cache_read_input_token_cost": 1.25e-07, "cache_read_input_token_cost_above_200k_tokens": 2.5e-07, - "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, + "deprecation_date": "2026-10-20", "input_cost_per_token": 1.25e-06, "input_cost_per_token_above_200k_tokens": 2.5e-06, "litellm_provider": "vertex_ai-language-models", @@ -20455,6 +20473,11 @@ "mode": "chat", "output_cost_per_token": 1e-05, "output_cost_per_token_above_200k_tokens": 1.5e-05, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -20479,21 +20502,20 @@ "supports_tool_choice": true, "supports_video_input": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "supports_web_search": true }, "gemini-3-pro-preview": { - "deprecation_date": "2026-03-26", + "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, + "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, + "cache_read_input_token_cost_priority": 3.6e-07, + "deprecation_date": "2026-03-26", "input_cost_per_token": 2e-06, "input_cost_per_token_above_200k_tokens": 4e-06, + "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, "input_cost_per_token_batches": 1e-06, + "input_cost_per_token_priority": 3.6e-06, "litellm_provider": "vertex_ai-language-models", "max_input_tokens": 1048576, "max_output_tokens": 65535, @@ -20501,7 +20523,14 @@ "mode": "chat", "output_cost_per_token": 1.2e-05, "output_cost_per_token_above_200k_tokens": 1.8e-05, + "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, "output_cost_per_token_batches": 6e-06, + "output_cost_per_token_priority": 2.16e-05, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -20519,6 +20548,7 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -20528,37 +20558,36 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 3.6e-06, - "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, - "output_cost_per_token_priority": 2.16e-05, - "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, - "cache_read_input_token_cost_priority": 3.6e-07, - "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini-3.1-pro-preview": { - "prompt_cache_min_tokens": 4096, + "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, + "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, + "cache_read_input_token_cost_priority": 3.6e-07, "input_cost_per_token": 2e-06, "input_cost_per_token_above_200k_tokens": 4e-06, + "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, "input_cost_per_token_batches": 1e-06, + "input_cost_per_token_priority": 3.6e-06, "litellm_provider": "vertex_ai-language-models", "max_input_tokens": 1048576, "max_output_tokens": 65536, "max_tokens": 65536, "mode": "chat", + "output_cost_per_image": 0.00012, "output_cost_per_token": 1.2e-05, "output_cost_per_token_above_200k_tokens": 1.8e-05, + "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, "output_cost_per_token_batches": 6e-06, - "output_cost_per_image": 0.00012, + "output_cost_per_token_priority": 2.16e-05, + "prompt_cache_min_tokens": 4096, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models", "supported_endpoints": [ "/v1/chat/completions", @@ -20576,35 +20605,23 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, + "supports_url_context": true, "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_url_context": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 3.6e-06, - "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, - "output_cost_per_token_priority": 2.16e-05, - "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, - "cache_read_input_token_cost_priority": 3.6e-07, - "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini-3.1-pro-preview-customtools": { - "prompt_cache_min_tokens": 4096, + "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "input_cost_per_token": 2e-06, "input_cost_per_token_above_200k_tokens": 4e-06, "input_cost_per_token_batches": 1e-06, @@ -20613,10 +20630,16 @@ "max_output_tokens": 65536, "max_tokens": 65536, "mode": "chat", + "output_cost_per_image": 0.00012, "output_cost_per_token": 1.2e-05, "output_cost_per_token_above_200k_tokens": 1.8e-05, "output_cost_per_token_batches": 6e-06, - "output_cost_per_image": 0.00012, + "prompt_cache_min_tokens": 4096, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models", "supported_endpoints": [ "/v1/chat/completions", @@ -20634,31 +20657,30 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, + "supports_url_context": true, "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_url_context": true, - "supports_native_streaming": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "vertex_ai/gemini-3-pro-preview": { + "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, + "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, + "cache_read_input_token_cost_priority": 3.6e-07, "input_cost_per_token": 2e-06, "input_cost_per_token_above_200k_tokens": 4e-06, + "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, "input_cost_per_token_batches": 1e-06, + "input_cost_per_token_priority": 3.6e-06, "litellm_provider": "vertex_ai", "max_input_tokens": 1048576, "max_output_tokens": 65535, @@ -20666,7 +20688,14 @@ "mode": "chat", "output_cost_per_token": 1.2e-05, "output_cost_per_token_above_200k_tokens": 1.8e-05, + "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, "output_cost_per_token_batches": 6e-06, + "output_cost_per_token_priority": 2.16e-05, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -20684,6 +20713,7 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -20693,30 +20723,27 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 3.6e-06, - "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, - "output_cost_per_token_priority": 2.16e-05, - "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, - "cache_read_input_token_cost_priority": 3.6e-07, - "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "vertex_ai/gemini-3-flash-preview": { "cache_read_input_token_cost": 5e-08, - "input_cost_per_token": 5e-07, + "cache_read_input_token_cost_priority": 9e-08, "input_cost_per_audio_token": 1e-06, + "input_cost_per_audio_token_priority": 1.8e-06, + "input_cost_per_token": 5e-07, + "input_cost_per_token_priority": 9e-07, "litellm_provider": "vertex_ai", "max_input_tokens": 1048576, "max_output_tokens": 65535, "max_tokens": 65535, "mode": "chat", "output_cost_per_token": 3e-06, + "output_cost_per_token_priority": 5.4e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -20734,6 +20761,7 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -20743,24 +20771,19 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 9e-07, - "input_cost_per_audio_token_priority": 1.8e-06, - "output_cost_per_token_priority": 5.4e-06, - "cache_read_input_token_cost_priority": 9e-08, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "vertex_ai/gemini-3.5-flash": { - "prompt_cache_min_tokens": 4096, - "deprecation_date": "2027-05-19", "cache_read_input_token_cost": 1.5e-07, - "input_cost_per_token": 1.5e-06, + "cache_read_input_token_cost_flex": 7.5e-08, + "cache_read_input_token_cost_priority": 2.7e-07, + "deprecation_date": "2027-05-19", "input_cost_per_audio_token": 1.5e-06, + "input_cost_per_audio_token_priority": 2.7e-06, + "input_cost_per_token": 1.5e-06, + "input_cost_per_token_batches": 7.5e-07, + "input_cost_per_token_flex": 7.5e-07, + "input_cost_per_token_priority": 2.7e-06, "litellm_provider": "vertex_ai", "max_input_tokens": 1048576, "max_output_tokens": 65535, @@ -20768,7 +20791,16 @@ "mode": "chat", "output_cost_per_reasoning_token": 9e-06, "output_cost_per_token": 9e-06, + "output_cost_per_token_batches": 4.5e-06, + "output_cost_per_token_flex": 4.5e-06, + "output_cost_per_token_priority": 1.62e-05, + "prompt_cache_min_tokens": 4096, "regional_endpoint_uplift_multiplier": 1.1, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -20786,6 +20818,7 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -20797,30 +20830,16 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 2.7e-06, - "input_cost_per_audio_token_priority": 2.7e-06, - "output_cost_per_token_priority": 1.62e-05, - "cache_read_input_token_cost_priority": 2.7e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, - "web_search_billing_unit": "per_query", - "input_cost_per_token_batches": 7.5e-07, - "output_cost_per_token_batches": 4.5e-06, - "input_cost_per_token_flex": 7.5e-07, - "output_cost_per_token_flex": 4.5e-06, - "cache_read_input_token_cost_flex": 7.5e-08 + "web_search_billing_unit": "per_query" }, "vertex_ai/gemini-3.6-flash": { - "prompt_cache_min_tokens": 4096, "cache_read_input_token_cost": 7.5e-08, "cache_read_input_token_cost_flex": 3.75e-08, + "cache_read_input_token_cost_priority": 1.35e-07, "input_cost_per_token": 7.5e-07, "input_cost_per_token_batches": 3.75e-07, "input_cost_per_token_flex": 3.75e-07, + "input_cost_per_token_priority": 1.35e-06, "litellm_provider": "vertex_ai", "max_input_tokens": 1048576, "max_output_tokens": 65536, @@ -20830,7 +20849,14 @@ "output_cost_per_token": 3.75e-06, "output_cost_per_token_batches": 1.875e-06, "output_cost_per_token_flex": 1.875e-06, + "output_cost_per_token_priority": 6.75e-06, + "prompt_cache_min_tokens": 4096, "regional_endpoint_uplift_multiplier": 1.1, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -20848,6 +20874,7 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -20859,24 +20886,16 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 1.35e-06, - "output_cost_per_token_priority": 6.75e-06, - "cache_read_input_token_cost_priority": 1.35e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "vertex_ai/gemini-3.7-flash": { - "prompt_cache_min_tokens": 4096, "cache_read_input_token_cost": 7.5e-08, "cache_read_input_token_cost_flex": 3.75e-08, + "cache_read_input_token_cost_priority": 1.35e-07, "input_cost_per_token": 7.5e-07, "input_cost_per_token_batches": 3.75e-07, "input_cost_per_token_flex": 3.75e-07, + "input_cost_per_token_priority": 1.35e-06, "litellm_provider": "vertex_ai", "max_input_tokens": 1048576, "max_output_tokens": 65536, @@ -20886,7 +20905,14 @@ "output_cost_per_token": 3.75e-06, "output_cost_per_token_batches": 1.875e-06, "output_cost_per_token_flex": 1.875e-06, + "output_cost_per_token_priority": 6.75e-06, + "prompt_cache_min_tokens": 4096, "regional_endpoint_uplift_multiplier": 1.1, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -20904,6 +20930,7 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -20915,34 +20942,36 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 1.35e-06, - "output_cost_per_token_priority": 6.75e-06, - "cache_read_input_token_cost_priority": 1.35e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "vertex_ai/gemini-3.1-pro-preview": { - "prompt_cache_min_tokens": 4096, + "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, + "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, + "cache_read_input_token_cost_priority": 3.6e-07, "input_cost_per_token": 2e-06, "input_cost_per_token_above_200k_tokens": 4e-06, + "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, "input_cost_per_token_batches": 1e-06, + "input_cost_per_token_priority": 3.6e-06, "litellm_provider": "vertex_ai", "max_input_tokens": 1048576, "max_output_tokens": 65536, "max_tokens": 65536, "mode": "chat", + "output_cost_per_image": 0.00012, "output_cost_per_token": 1.2e-05, "output_cost_per_token_above_200k_tokens": 1.8e-05, + "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, "output_cost_per_token_batches": 6e-06, - "output_cost_per_image": 0.00012, + "output_cost_per_token_priority": 2.16e-05, + "prompt_cache_min_tokens": 4096, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models", "supported_endpoints": [ "/v1/chat/completions", @@ -20960,47 +20989,47 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, + "supports_url_context": true, "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_url_context": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 3.6e-06, - "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, - "output_cost_per_token_priority": 2.16e-05, - "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, - "cache_read_input_token_cost_priority": 3.6e-07, - "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "vertex_ai/gemini-3.1-pro-preview-customtools": { - "prompt_cache_min_tokens": 4096, + "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, + "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, + "cache_read_input_token_cost_priority": 3.6e-07, "input_cost_per_token": 2e-06, "input_cost_per_token_above_200k_tokens": 4e-06, + "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, "input_cost_per_token_batches": 1e-06, + "input_cost_per_token_priority": 3.6e-06, "litellm_provider": "vertex_ai", "max_input_tokens": 1048576, "max_output_tokens": 65536, "max_tokens": 65536, "mode": "chat", + "output_cost_per_image": 0.00012, "output_cost_per_token": 1.2e-05, "output_cost_per_token_above_200k_tokens": 1.8e-05, + "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, "output_cost_per_token_batches": 6e-06, - "output_cost_per_image": 0.00012, + "output_cost_per_token_priority": 2.16e-05, + "prompt_cache_min_tokens": 4096, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models", "supported_endpoints": [ "/v1/chat/completions", @@ -21018,28 +21047,17 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, + "supports_url_context": true, "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_url_context": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 3.6e-06, - "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, - "output_cost_per_token_priority": 2.16e-05, - "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, - "cache_read_input_token_cost_priority": 3.6e-07, - "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini-2.5-pro-preview-tts": { @@ -21055,6 +21073,11 @@ "mode": "chat", "output_cost_per_token": 1e-05, "output_cost_per_token_above_200k_tokens": 1.5e-05, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-2.5-pro-preview", "supported_modalities": [ "text" @@ -21070,24 +21093,19 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "supports_web_search": true }, "gemini-robotics-er-1.5-preview": { "cache_read_input_token_cost": 0, - "input_cost_per_token": 3e-07, "input_cost_per_audio_token": 1e-06, + "input_cost_per_token": 3e-07, "litellm_provider": "vertex_ai-language-models", "max_input_tokens": 1048576, "max_output_tokens": 65535, "max_tokens": 65535, "mode": "chat", - "output_cost_per_token": 2.5e-06, "output_cost_per_reasoning_token": 2.5e-06, + "output_cost_per_token": 2.5e-06, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-robotics-er-1-5-preview", "supported_endpoints": [ "/v1/chat/completions", @@ -21116,15 +21134,21 @@ "gemini/gemini-robotics-er-1.5-preview": { "cache_read_input_token_cost": 0, "deprecation_date": "2026-04-30", - "input_cost_per_token": 3e-07, "input_cost_per_audio_token": 1e-06, + "input_cost_per_token": 3e-07, "litellm_provider": "gemini", "max_input_tokens": 1048576, "max_output_tokens": 65535, "max_tokens": 65535, "mode": "chat", - "output_cost_per_token": 2.5e-06, "output_cost_per_reasoning_token": 2.5e-06, + "output_cost_per_token": 2.5e-06, + "rpm": 10, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-robotics-er-1-5-preview", "supported_endpoints": [ "/v1/chat/completions", @@ -21150,13 +21174,7 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 250000, - "rpm": 10, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 250000 }, "gemini/gemini-robotics-er-2-preview": { "cache_read_input_token_cost": 2e-07, @@ -21432,6 +21450,11 @@ "mode": "chat", "output_cost_per_token": 4e-07, "rpm": 10000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/pricing#2_0flash", "supported_modalities": [ "text", @@ -21453,12 +21476,7 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 10000000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 10000000 }, "gemini/gemini-2.0-flash-001": { "cache_read_input_token_cost": 2.5e-08, @@ -21472,6 +21490,11 @@ "mode": "chat", "output_cost_per_token": 4e-07, "rpm": 10000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/pricing#2_0flash", "supported_modalities": [ "text", @@ -21491,12 +21514,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "tpm": 10000000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 10000000 }, "gemini/gemini-2.0-flash-lite": { "cache_read_input_token_cost": 1.875e-08, @@ -21509,6 +21527,11 @@ "mode": "chat", "output_cost_per_token": 3e-07, "rpm": 4000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-2.0-flash-lite", "supported_modalities": [ "text", @@ -21527,12 +21550,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "tpm": 4000000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 4000000 }, "gemini/gemini-2.5-flash": { "cache_read_input_token_cost": 3e-08, @@ -21546,6 +21564,11 @@ "output_cost_per_reasoning_token": 2.5e-06, "output_cost_per_token": 2.5e-06, "rpm": 100000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-preview", "supported_endpoints": [ "/v1/chat/completions", @@ -21563,6 +21586,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -21573,13 +21597,7 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 8000000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "supports_image_size": false + "tpm": 8000000 }, "gemini/gemini-2.5-flash-image": { "cache_read_input_token_cost": 3e-08, @@ -21587,7 +21605,6 @@ "input_cost_per_audio_token": 1e-06, "input_cost_per_token": 3e-07, "litellm_provider": "gemini", - "supports_reasoning": false, "max_input_tokens": 32768, "max_output_tokens": 32768, "max_tokens": 32768, @@ -21597,6 +21614,11 @@ "output_cost_per_reasoning_token": 2.5e-06, "output_cost_per_token": 2.5e-06, "rpm": 100000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-2.5-flash-image", "supported_endpoints": [ "/v1/chat/completions", @@ -21615,22 +21637,18 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, + "supports_reasoning": false, "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 8000000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "supports_image_size": false + "tpm": 8000000 }, "gemini/gemini-3-pro-image": { "input_cost_per_image": 0.0011, @@ -21644,9 +21662,13 @@ "output_cost_per_image": 0.134, "output_cost_per_image_token": 0.00012, "output_cost_per_token": 1.2e-05, - "rpm": 1000, - "tpm": 4000000, "output_cost_per_token_batches": 6e-06, + "rpm": 1000, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-3-pro-image", "supported_endpoints": [ "/v1/chat/completions", @@ -21663,18 +21685,14 @@ ], "supports_function_calling": false, "supports_prompt_caching": true, + "supports_reasoning": false, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_vision": true, "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, - "web_search_billing_unit": "per_query", - "supports_reasoning": false, - "supports_service_tier": true + "tpm": 4000000, + "web_search_billing_unit": "per_query" }, "gemini/gemini-3-pro-image-preview": { "deprecation_date": "2026-06-25", @@ -21689,9 +21707,13 @@ "output_cost_per_image": 0.134, "output_cost_per_image_token": 0.00012, "output_cost_per_token": 1.2e-05, - "rpm": 1000, - "tpm": 4000000, "output_cost_per_token_batches": 6e-06, + "rpm": 1000, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -21713,11 +21735,7 @@ "supports_system_messages": true, "supports_vision": true, "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, + "tpm": 4000000, "web_search_billing_unit": "per_query" }, "gemini/gemini-3.1-flash-image": { @@ -21730,10 +21748,15 @@ "mode": "image_generation", "output_cost_per_image": 0.045, "output_cost_per_image_token": 6e-05, + "output_cost_per_image_token_batches": 3e-05, "output_cost_per_token": 3e-06, "output_cost_per_token_batches": 1.5e-06, "rpm": 1000, - "tpm": 4000000, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-3.1-flash-image", "supported_endpoints": [ "/v1/chat/completions", @@ -21755,13 +21778,8 @@ "supports_system_messages": true, "supports_vision": true, "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, - "web_search_billing_unit": "per_query", - "output_cost_per_image_token_batches": 3e-05 + "tpm": 4000000, + "web_search_billing_unit": "per_query" }, "gemini/gemini-3.1-flash-image-preview": { "deprecation_date": "2026-06-25", @@ -21777,7 +21795,11 @@ "output_cost_per_token": 3e-06, "output_cost_per_token_batches": 1.5e-06, "rpm": 1000, - "tpm": 4000000, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-3.1-flash-image-preview", "supported_endpoints": [ "/v1/chat/completions", @@ -21799,11 +21821,7 @@ "supports_system_messages": true, "supports_vision": true, "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, + "tpm": 4000000, "web_search_billing_unit": "per_query" }, "gemini/gemini-3.1-flash-lite-image": { @@ -21854,9 +21872,13 @@ "output_cost_per_image": 0.134, "output_cost_per_image_token": 0.00012, "output_cost_per_token": 1.2e-05, - "rpm": 1000, - "tpm": 4000000, "output_cost_per_token_batches": 6e-06, + "rpm": 1000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -21877,11 +21899,7 @@ "supports_system_messages": true, "supports_vision": true, "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 4000000 }, "gemini/gemini-2.5-flash-lite": { "cache_read_input_token_cost": 1e-08, @@ -21895,6 +21913,11 @@ "output_cost_per_reasoning_token": 4e-07, "output_cost_per_token": 4e-07, "rpm": 15, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-lite", "supported_endpoints": [ "/v1/chat/completions", @@ -21912,6 +21935,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -21922,13 +21946,7 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 250000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "supports_image_size": false + "tpm": 250000 }, "gemini/gemini-2.5-flash-lite-preview-09-2025": { "cache_read_input_token_cost": 1e-08, @@ -21943,6 +21961,11 @@ "output_cost_per_reasoning_token": 4e-07, "output_cost_per_token": 4e-07, "rpm": 15, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://developers.googleblog.com/en/continuing-to-bring-you-our-latest-models-with-an-improved-gemini-2-5-flash-and-flash-lite-release/", "supported_endpoints": [ "/v1/chat/completions", @@ -21960,6 +21983,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -21970,13 +21994,7 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 250000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "supports_image_size": false + "tpm": 250000 }, "gemini/gemini-2.5-flash-preview-09-2025": { "cache_read_input_token_cost": 7.5e-08, @@ -21991,6 +22009,11 @@ "output_cost_per_reasoning_token": 2.5e-06, "output_cost_per_token": 2.5e-06, "rpm": 15, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://developers.googleblog.com/en/continuing-to-bring-you-our-latest-models-with-an-improved-gemini-2-5-flash-and-flash-lite-release/", "supported_endpoints": [ "/v1/chat/completions", @@ -22008,6 +22031,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -22018,13 +22042,7 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 250000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "supports_image_size": false + "tpm": 250000 }, "gemini/gemini-flash-latest": { "cache_read_input_token_cost": 7.5e-08, @@ -22038,6 +22056,11 @@ "output_cost_per_reasoning_token": 2.5e-06, "output_cost_per_token": 2.5e-06, "rpm": 15, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://developers.googleblog.com/en/continuing-to-bring-you-our-latest-models-with-an-improved-gemini-2-5-flash-and-flash-lite-release/", "supported_endpoints": [ "/v1/chat/completions", @@ -22065,12 +22088,7 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 250000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 250000 }, "gemini/gemini-flash-lite-latest": { "cache_read_input_token_cost": 2.5e-08, @@ -22084,6 +22102,11 @@ "output_cost_per_reasoning_token": 4e-07, "output_cost_per_token": 4e-07, "rpm": 15, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://developers.googleblog.com/en/continuing-to-bring-you-our-latest-models-with-an-improved-gemini-2-5-flash-and-flash-lite-release/", "supported_endpoints": [ "/v1/chat/completions", @@ -22111,16 +22134,11 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 250000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 250000 }, "gemini/gemini-2.5-flash-lite-preview-06-17": { - "deprecation_date": "2025-11-18", "cache_read_input_token_cost": 2.5e-08, + "deprecation_date": "2025-11-18", "input_cost_per_audio_token": 5e-07, "input_cost_per_token": 1e-07, "litellm_provider": "gemini", @@ -22131,6 +22149,11 @@ "output_cost_per_reasoning_token": 4e-07, "output_cost_per_token": 4e-07, "rpm": 15, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-lite", "supported_endpoints": [ "/v1/chat/completions", @@ -22148,6 +22171,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -22158,33 +22182,27 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 250000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "supports_image_size": false + "tpm": 250000 }, "gemini/gemini-2.5-flash-preview-tts": { "input_cost_per_token": 5e-07, "litellm_provider": "gemini", "mode": "audio_speech", "output_cost_per_token": 1e-05, + "rpm": 10, "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_endpoints": [ "/v1/audio/speech" ], - "tpm": 4000000, - "rpm": 10 + "tpm": 4000000 }, "gemini/gemini-2.5-pro": { "cache_read_input_token_cost": 1.25e-07, "cache_read_input_token_cost_above_200k_tokens": 2.5e-07, "input_cost_per_token": 1.25e-06, "input_cost_per_token_above_200k_tokens": 2.5e-06, - "input_cost_per_token_priority": 1.25e-06, "input_cost_per_token_above_200k_tokens_priority": 2.5e-06, + "input_cost_per_token_priority": 1.25e-06, "litellm_provider": "gemini", "max_input_tokens": 1048576, "max_output_tokens": 65535, @@ -22192,9 +22210,14 @@ "mode": "chat", "output_cost_per_token": 1e-05, "output_cost_per_token_above_200k_tokens": 1.5e-05, - "output_cost_per_token_priority": 1e-05, "output_cost_per_token_above_200k_tokens_priority": 1.5e-05, + "output_cost_per_token_priority": 1e-05, "rpm": 2000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -22220,12 +22243,7 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "tpm": 800000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 800000 }, "gemini/gemini-2.5-computer-use-preview-10-2025": { "input_cost_per_token": 1.25e-06, @@ -22258,12 +22276,16 @@ "tpm": 800000 }, "gemini/gemini-3-pro-preview": { - "deprecation_date": "2026-03-09", "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, + "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, + "cache_read_input_token_cost_priority": 3.6e-07, + "deprecation_date": "2026-03-09", "input_cost_per_token": 2e-06, "input_cost_per_token_above_200k_tokens": 4e-06, + "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, "input_cost_per_token_batches": 1e-06, + "input_cost_per_token_priority": 3.6e-06, "litellm_provider": "gemini", "max_input_tokens": 1048576, "max_output_tokens": 65535, @@ -22271,8 +22293,15 @@ "mode": "chat", "output_cost_per_token": 1.2e-05, "output_cost_per_token_above_200k_tokens": 1.8e-05, + "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, "output_cost_per_token_batches": 6e-06, + "output_cost_per_token_priority": 2.16e-05, "rpm": 2000, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -22300,17 +22329,6 @@ "supports_vision": true, "supports_web_search": true, "tpm": 800000, - "input_cost_per_token_priority": 3.6e-06, - "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, - "output_cost_per_token_priority": 2.16e-05, - "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, - "cache_read_input_token_cost_priority": 3.6e-07, - "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini/gemini-3.1-flash-lite-preview": { @@ -22326,6 +22344,11 @@ "output_cost_per_reasoning_token": 1.5e-06, "output_cost_per_token": 1.5e-06, "rpm": 15, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/models", "supported_endpoints": [ "/v1/chat/completions", @@ -22344,6 +22367,7 @@ "supports_audio_input": true, "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -22355,18 +22379,13 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, "tpm": 250000, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini/gemini-3.1-flash-lite": { "cache_read_input_token_cost": 2.5e-08, "cache_read_input_token_cost_flex": 1.25e-08, + "cache_read_input_token_cost_per_audio_token": 5e-08, "cache_read_input_token_cost_priority": 4.5e-08, "deprecation_date": "2027-05-07", "input_cost_per_audio_token": 5e-07, @@ -22375,9 +22394,13 @@ "input_cost_per_token_flex": 1.25e-07, "input_cost_per_token_priority": 4.5e-07, "litellm_provider": "gemini", + "max_audio_length_hours": 8.4, + "max_audio_per_prompt": 1, + "max_images_per_prompt": 3000, "max_input_tokens": 1048576, "max_output_tokens": 65536, "max_tokens": 65536, + "max_videos_per_prompt": 10, "mode": "chat", "output_cost_per_reasoning_token": 1.5e-06, "output_cost_per_token": 1.5e-06, @@ -22385,6 +22408,11 @@ "output_cost_per_token_flex": 7.5e-07, "output_cost_per_token_priority": 2.7e-06, "rpm": 15, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-3.1-flash-lite", "supported_endpoints": [ "/v1/chat/completions", @@ -22402,34 +22430,24 @@ ], "supports_audio_input": true, "supports_audio_output": false, + "supports_code_execution": true, + "supports_file_search": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_tool_choice": true, "supports_url_context": true, "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, "tpm": 250000, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, - "web_search_billing_unit": "per_query", - "cache_read_input_token_cost_per_audio_token": 5e-08, - "max_audio_length_hours": 8.4, - "max_audio_per_prompt": 1, - "max_images_per_prompt": 3000, - "max_videos_per_prompt": 10, - "supports_code_execution": true, - "supports_file_search": true, - "supports_service_tier": true + "web_search_billing_unit": "per_query" }, "gemini/gemini-3.5-flash-lite": { "cache_read_input_token_cost": 3e-08, @@ -22450,6 +22468,11 @@ "output_cost_per_token_flex": 1.25e-06, "output_cost_per_token_priority": 4.5e-06, "rpm": 15, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -22468,6 +22491,7 @@ "supports_audio_input": true, "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -22479,19 +22503,16 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, "tpm": 250000, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini/gemini-3-flash-preview": { "cache_read_input_token_cost": 5e-08, + "cache_read_input_token_cost_priority": 9e-08, "input_cost_per_audio_token": 1e-06, + "input_cost_per_audio_token_priority": 1.8e-06, "input_cost_per_token": 5e-07, + "input_cost_per_token_priority": 9e-07, "litellm_provider": "gemini", "max_input_tokens": 1048576, "max_output_tokens": 65535, @@ -22499,7 +22520,13 @@ "mode": "chat", "output_cost_per_reasoning_token": 3e-06, "output_cost_per_token": 3e-06, + "output_cost_per_token_priority": 5.4e-06, "rpm": 2000, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/pricing/gemini-3", "supported_endpoints": [ "/v1/chat/completions", @@ -22517,6 +22544,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -22527,24 +22555,19 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, "tpm": 800000, - "input_cost_per_token_priority": 9e-07, - "input_cost_per_audio_token_priority": 1.8e-06, - "output_cost_per_token_priority": 5.4e-06, - "cache_read_input_token_cost_priority": 9e-08, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini/gemini-3.5-flash": { - "prompt_cache_min_tokens": 4096, "cache_read_input_token_cost": 1.5e-07, + "cache_read_input_token_cost_flex": 8e-08, + "cache_read_input_token_cost_priority": 2.7e-07, "input_cost_per_audio_token": 1.5e-06, + "input_cost_per_audio_token_priority": 2.7e-06, "input_cost_per_token": 1.5e-06, + "input_cost_per_token_batches": 7.5e-07, + "input_cost_per_token_flex": 7.5e-07, + "input_cost_per_token_priority": 2.7e-06, "litellm_provider": "gemini", "max_input_tokens": 1048576, "max_output_tokens": 65535, @@ -22552,7 +22575,16 @@ "mode": "chat", "output_cost_per_reasoning_token": 9e-06, "output_cost_per_token": 9e-06, + "output_cost_per_token_batches": 4.5e-06, + "output_cost_per_token_flex": 4.5e-06, + "output_cost_per_token_priority": 1.62e-05, + "prompt_cache_min_tokens": 4096, "rpm": 2000, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/pricing/gemini-3", "supported_endpoints": [ "/v1/chat/completions", @@ -22568,9 +22600,10 @@ "supported_output_modalities": [ "text" ], - "supports_audio_output": false, "supports_audio_input": true, + "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -22582,31 +22615,17 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, "tpm": 800000, - "input_cost_per_token_priority": 2.7e-06, - "input_cost_per_audio_token_priority": 2.7e-06, - "output_cost_per_token_priority": 1.62e-05, - "cache_read_input_token_cost_priority": 2.7e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, - "web_search_billing_unit": "per_query", - "input_cost_per_token_batches": 7.5e-07, - "output_cost_per_token_batches": 4.5e-06, - "input_cost_per_token_flex": 7.5e-07, - "output_cost_per_token_flex": 4.5e-06, - "cache_read_input_token_cost_flex": 8e-08 + "web_search_billing_unit": "per_query" }, "gemini/gemini-3.6-flash": { - "prompt_cache_min_tokens": 4096, "cache_read_input_token_cost": 7.5e-08, "cache_read_input_token_cost_flex": 3.75e-08, + "cache_read_input_token_cost_priority": 1.35e-07, "input_cost_per_token": 7.5e-07, "input_cost_per_token_batches": 3.75e-07, "input_cost_per_token_flex": 3.75e-07, + "input_cost_per_token_priority": 1.35e-06, "litellm_provider": "gemini", "max_input_tokens": 1048576, "max_output_tokens": 65536, @@ -22616,7 +22635,14 @@ "output_cost_per_token": 3.75e-06, "output_cost_per_token_batches": 1.875e-06, "output_cost_per_token_flex": 1.875e-06, + "output_cost_per_token_priority": 6.75e-06, + "prompt_cache_min_tokens": 4096, "rpm": 2000, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/pricing/gemini-3", "supported_endpoints": [ "/v1/chat/completions", @@ -22632,9 +22658,10 @@ "supported_output_modalities": [ "text" ], - "supports_audio_output": false, "supports_audio_input": true, + "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -22646,25 +22673,17 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, "tpm": 800000, - "input_cost_per_token_priority": 1.35e-06, - "output_cost_per_token_priority": 6.75e-06, - "cache_read_input_token_cost_priority": 1.35e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini/gemini-3.7-flash": { - "prompt_cache_min_tokens": 4096, "cache_read_input_token_cost": 7.5e-08, "cache_read_input_token_cost_flex": 3.75e-08, + "cache_read_input_token_cost_priority": 1.35e-07, "input_cost_per_token": 7.5e-07, "input_cost_per_token_batches": 3.75e-07, "input_cost_per_token_flex": 3.75e-07, + "input_cost_per_token_priority": 1.35e-06, "litellm_provider": "gemini", "max_input_tokens": 1048576, "max_output_tokens": 65536, @@ -22674,7 +22693,14 @@ "output_cost_per_token": 3.75e-06, "output_cost_per_token_batches": 1.875e-06, "output_cost_per_token_flex": 1.875e-06, + "output_cost_per_token_priority": 6.75e-06, + "prompt_cache_min_tokens": 4096, "rpm": 2000, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/pricing/gemini-3", "supported_endpoints": [ "/v1/chat/completions", @@ -22690,9 +22716,10 @@ "supported_output_modalities": [ "text" ], - "supports_audio_output": false, "supports_audio_input": true, + "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -22704,16 +22731,7 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, "tpm": 800000, - "input_cost_per_token_priority": 1.35e-06, - "output_cost_per_token_priority": 6.75e-06, - "cache_read_input_token_cost_priority": 1.35e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini/gemini-omni-flash-preview": { @@ -22750,12 +22768,15 @@ "tpm": 800000 }, "gemini/gemini-3.1-pro-preview": { - "prompt_cache_min_tokens": 4096, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, + "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, + "cache_read_input_token_cost_priority": 3.6e-07, "input_cost_per_token": 2e-06, "input_cost_per_token_above_200k_tokens": 4e-06, + "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, "input_cost_per_token_batches": 1e-06, + "input_cost_per_token_priority": 3.6e-06, "litellm_provider": "gemini", "max_input_tokens": 1048576, "max_output_tokens": 65536, @@ -22763,8 +22784,16 @@ "mode": "chat", "output_cost_per_token": 1.2e-05, "output_cost_per_token_above_200k_tokens": 1.8e-05, + "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, "output_cost_per_token_batches": 6e-06, + "output_cost_per_token_priority": 2.16e-05, + "prompt_cache_min_tokens": 4096, "rpm": 2000, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-3.1-pro-preview", "supported_endpoints": [ "/v1/chat/completions", @@ -22782,38 +22811,30 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, + "supports_url_context": true, "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_url_context": true, - "supports_native_streaming": true, "tpm": 800000, - "input_cost_per_token_priority": 3.6e-06, - "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, - "output_cost_per_token_priority": 2.16e-05, - "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, - "cache_read_input_token_cost_priority": 3.6e-07, - "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini/gemini-3.1-pro-preview-customtools": { - "prompt_cache_min_tokens": 4096, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, + "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, + "cache_read_input_token_cost_priority": 3.6e-07, "input_cost_per_token": 2e-06, "input_cost_per_token_above_200k_tokens": 4e-06, + "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, "input_cost_per_token_batches": 1e-06, + "input_cost_per_token_priority": 3.6e-06, "litellm_provider": "gemini", "max_input_tokens": 1048576, "max_output_tokens": 65536, @@ -22821,8 +22842,16 @@ "mode": "chat", "output_cost_per_token": 1.2e-05, "output_cost_per_token_above_200k_tokens": 1.8e-05, + "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, "output_cost_per_token_batches": 6e-06, + "output_cost_per_token_priority": 2.16e-05, + "prompt_cache_min_tokens": 4096, "rpm": 2000, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-3.1-pro-preview", "supported_endpoints": [ "/v1/chat/completions", @@ -22840,35 +22869,27 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, + "supports_url_context": true, "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_url_context": true, - "supports_native_streaming": true, "tpm": 800000, - "input_cost_per_token_priority": 3.6e-06, - "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, - "output_cost_per_token_priority": 2.16e-05, - "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, - "cache_read_input_token_cost_priority": 3.6e-07, - "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini-3-flash-preview": { "cache_read_input_token_cost": 5e-08, + "cache_read_input_token_cost_priority": 9e-08, "input_cost_per_audio_token": 1e-06, + "input_cost_per_audio_token_priority": 1.8e-06, "input_cost_per_token": 5e-07, + "input_cost_per_token_priority": 9e-07, "litellm_provider": "vertex_ai-language-models", "max_input_tokens": 1048576, "max_output_tokens": 65535, @@ -22876,6 +22897,12 @@ "mode": "chat", "output_cost_per_reasoning_token": 3e-06, "output_cost_per_token": 3e-06, + "output_cost_per_token_priority": 5.4e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/pricing/gemini-3", "supported_endpoints": [ "/v1/chat/completions", @@ -22893,6 +22920,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -22903,16 +22931,6 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 9e-07, - "input_cost_per_audio_token_priority": 1.8e-06, - "output_cost_per_token_priority": 5.4e-06, - "cache_read_input_token_cost_priority": 9e-08, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini-omni-flash-preview": { @@ -22947,11 +22965,16 @@ "supports_vision": true }, "gemini-3.5-flash": { - "prompt_cache_min_tokens": 4096, - "deprecation_date": "2027-05-19", "cache_read_input_token_cost": 1.5e-07, + "cache_read_input_token_cost_flex": 7.5e-08, + "cache_read_input_token_cost_priority": 2.7e-07, + "deprecation_date": "2027-05-19", "input_cost_per_audio_token": 1.5e-06, + "input_cost_per_audio_token_priority": 2.7e-06, "input_cost_per_token": 1.5e-06, + "input_cost_per_token_batches": 7.5e-07, + "input_cost_per_token_flex": 7.5e-07, + "input_cost_per_token_priority": 2.7e-06, "litellm_provider": "vertex_ai-language-models", "max_input_tokens": 1048576, "max_output_tokens": 65535, @@ -22959,6 +22982,15 @@ "mode": "chat", "output_cost_per_reasoning_token": 9e-06, "output_cost_per_token": 9e-06, + "output_cost_per_token_batches": 4.5e-06, + "output_cost_per_token_flex": 4.5e-06, + "output_cost_per_token_priority": 1.62e-05, + "prompt_cache_min_tokens": 4096, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/pricing/gemini-3", "supported_endpoints": [ "/v1/chat/completions", @@ -22974,9 +23006,10 @@ "supported_output_modalities": [ "text" ], - "supports_audio_output": false, "supports_audio_input": true, + "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -22988,30 +23021,16 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 2.7e-06, - "input_cost_per_audio_token_priority": 2.7e-06, - "output_cost_per_token_priority": 1.62e-05, - "cache_read_input_token_cost_priority": 2.7e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, - "web_search_billing_unit": "per_query", - "input_cost_per_token_batches": 7.5e-07, - "output_cost_per_token_batches": 4.5e-06, - "input_cost_per_token_flex": 7.5e-07, - "output_cost_per_token_flex": 4.5e-06, - "cache_read_input_token_cost_flex": 7.5e-08 + "web_search_billing_unit": "per_query" }, "gemini-3.6-flash": { - "prompt_cache_min_tokens": 4096, "cache_read_input_token_cost": 7.5e-08, "cache_read_input_token_cost_flex": 3.75e-08, + "cache_read_input_token_cost_priority": 1.35e-07, "input_cost_per_token": 7.5e-07, "input_cost_per_token_batches": 3.75e-07, "input_cost_per_token_flex": 3.75e-07, + "input_cost_per_token_priority": 1.35e-06, "litellm_provider": "vertex_ai-language-models", "max_input_tokens": 1048576, "max_output_tokens": 65536, @@ -23021,6 +23040,13 @@ "output_cost_per_token": 3.75e-06, "output_cost_per_token_batches": 1.875e-06, "output_cost_per_token_flex": 1.875e-06, + "output_cost_per_token_priority": 6.75e-06, + "prompt_cache_min_tokens": 4096, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/pricing/gemini-3", "supported_endpoints": [ "/v1/chat/completions", @@ -23036,9 +23062,10 @@ "supported_output_modalities": [ "text" ], - "supports_audio_output": false, "supports_audio_input": true, + "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -23050,24 +23077,16 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 1.35e-06, - "output_cost_per_token_priority": 6.75e-06, - "cache_read_input_token_cost_priority": 1.35e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini-3.7-flash": { - "prompt_cache_min_tokens": 4096, "cache_read_input_token_cost": 7.5e-08, "cache_read_input_token_cost_flex": 3.75e-08, + "cache_read_input_token_cost_priority": 1.35e-07, "input_cost_per_token": 7.5e-07, "input_cost_per_token_batches": 3.75e-07, "input_cost_per_token_flex": 3.75e-07, + "input_cost_per_token_priority": 1.35e-06, "litellm_provider": "vertex_ai-language-models", "max_input_tokens": 1048576, "max_output_tokens": 65536, @@ -23077,6 +23096,13 @@ "output_cost_per_token": 3.75e-06, "output_cost_per_token_batches": 1.875e-06, "output_cost_per_token_flex": 1.875e-06, + "output_cost_per_token_priority": 6.75e-06, + "prompt_cache_min_tokens": 4096, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/pricing/gemini-3", "supported_endpoints": [ "/v1/chat/completions", @@ -23092,9 +23118,10 @@ "supported_output_modalities": [ "text" ], - "supports_audio_output": false, "supports_audio_input": true, + "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -23106,15 +23133,6 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 1.35e-06, - "output_cost_per_token_priority": 6.75e-06, - "cache_read_input_token_cost_priority": 1.35e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini/gemini-2.5-pro-preview-tts": { @@ -23131,6 +23149,11 @@ "output_cost_per_token": 1e-05, "output_cost_per_token_above_200k_tokens": 1.5e-05, "rpm": 10000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-2.5-pro-preview", "supported_modalities": [ "text" @@ -23146,12 +23169,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "tpm": 10000000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 10000000 }, "gemini/gemini-exp-1114": { "input_cost_per_token": 0, @@ -23206,12 +23224,12 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.05e-06, + "rpm": 10, "source": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/models#foundation_models", "supports_function_calling": true, "supports_tool_choice": true, "supports_vision": true, - "tpm": 250000, - "rpm": 10 + "tpm": 250000 }, "gemini/gemini-gemma-2-9b-it": { "input_cost_per_token": 3.5e-07, @@ -23220,12 +23238,12 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.05e-06, + "rpm": 10, "source": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/models#foundation_models", "supports_function_calling": true, "supports_tool_choice": true, "supports_vision": true, - "tpm": 250000, - "rpm": 10 + "tpm": 250000 }, "gemini/gemma-3-27b-it": { "input_cost_per_audio_per_second": 0, @@ -23459,22 +23477,22 @@ ] }, "github_copilot/claude-haiku-4.5": { + "cache_creation_input_token_cost": 1.25e-06, + "cache_read_input_token_cost": 1e-07, + "input_cost_per_token": 1e-06, "litellm_provider": "github_copilot", "max_input_tokens": 128000, "max_output_tokens": 16000, "max_tokens": 16000, "mode": "chat", + "output_cost_per_token": 5e-06, + "source": "https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing", "supported_endpoints": [ "/v1/chat/completions" ], "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, - "cache_creation_input_token_cost": 1.25e-06, - "cache_read_input_token_cost": 1e-07, - "input_cost_per_token": 1e-06, - "output_cost_per_token": 5e-06, - "source": "https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing" + "supports_vision": true }, "github_copilot/claude-opus-4.5": { "litellm_provider": "github_copilot", @@ -23486,9 +23504,24 @@ "/v1/chat/completions" ], "supports_function_calling": true, + "supports_output_config": true, "supports_parallel_function_calling": true, - "supports_vision": true, - "supports_output_config": true + "supports_vision": true + }, + "github_copilot/claude-opus-4.6-fast": { + "supports_adaptive_thinking": true, + "supports_legacy_thinking": true, + "litellm_provider": "github_copilot", + "max_input_tokens": 128000, + "max_output_tokens": 16000, + "max_tokens": 16000, + "mode": "chat", + "supported_endpoints": [ + "/v1/chat/completions" + ], + "supports_function_calling": true, + "supports_parallel_function_calling": true, + "supports_vision": true }, "github_copilot/claude-sonnet-4.5": { "litellm_provider": "github_copilot", @@ -23616,38 +23649,38 @@ "supports_parallel_function_calling": true }, "github_copilot/gpt-5-mini": { + "cache_read_input_token_cost": 2.5e-08, + "input_cost_per_token": 2.5e-07, "litellm_provider": "github_copilot", "max_input_tokens": 128000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", + "output_cost_per_token": 2e-06, + "source": "https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing", "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_response_schema": true, - "supports_vision": true, - "cache_read_input_token_cost": 2.5e-08, - "input_cost_per_token": 2.5e-07, - "output_cost_per_token": 2e-06, - "source": "https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing" + "supports_vision": true }, "github_copilot/gpt-5.3-codex": { + "cache_read_input_token_cost": 1.75e-07, + "input_cost_per_token": 1.75e-06, "litellm_provider": "github_copilot", "max_input_tokens": 128000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "responses", + "output_cost_per_token": 1.4e-05, + "source": "https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing", "supported_endpoints": [ "/v1/responses" ], "supports_function_calling": true, "supports_parallel_function_calling": true, + "supports_reasoning": true, "supports_response_schema": true, - "supports_vision": true, - "cache_read_input_token_cost": 1.75e-07, - "input_cost_per_token": 1.75e-06, - "output_cost_per_token": 1.4e-05, - "source": "https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing", - "supports_reasoning": true + "supports_vision": true }, "github_copilot/mai-code-1-flash": { "cache_read_input_token_cost": 7.5e-08, @@ -23914,8 +23947,8 @@ "mode": "chat", "output_cost_per_token": 2.5e-05, "supports_function_calling": true, - "supports_vision": true, - "supports_output_config": true + "supports_output_config": true, + "supports_vision": true }, "gmi/anthropic/claude-sonnet-4.5": { "input_cost_per_token": 3e-06, @@ -24188,20 +24221,22 @@ "global.anthropic.claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, + "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.2e-05, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.2e-05, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "input_cost_per_token_batches": 1.5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "output_cost_per_token_batches": 7.5e-06, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -24210,33 +24245,33 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024, - "input_cost_per_token_batches": 1.5e-06, - "output_cost_per_token_batches": 7.5e-06 + "supports_tool_search": true, + "supports_vision": true }, "global.anthropic.claude-sonnet-4-20250514-v1:0": { + "bedrock_converse_supports_strict_tools": false, "cache_creation_input_token_cost": 3.75e-06, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, "deprecation_date": "2026-10-14", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -24250,37 +24285,35 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "bedrock_converse_supports_strict_tools": false, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "global.anthropic.claude-haiku-4-5-20251001-v1:0": { "cache_creation_input_token_cost": 1.25e-06, "cache_creation_input_token_cost_above_1hr": 2e-06, "cache_read_input_token_cost": 1e-07, "input_cost_per_token": 1e-06, + "input_cost_per_token_batches": 5e-07, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, + "output_cost_per_token_batches": 2.5e-06, + "prompt_cache_min_tokens": 4096, "source": "https://aws.amazon.com/about-aws/whats-new/2025/10/claude-4-5-haiku-anthropic-amazon-bedrock", "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096, - "input_cost_per_token_batches": 5e-07, - "output_cost_per_token_batches": 2.5e-06 + "supports_tool_search": true, + "supports_vision": true }, "global.amazon.nova-2-lite-v1:0": { "cache_read_input_token_cost": 7.5e-08, @@ -24541,16 +24574,16 @@ "cache_read_input_token_cost": 5e-07, "cache_read_input_token_cost_priority": 8.75e-07, "input_cost_per_token": 2e-06, - "input_cost_per_token_priority": 3.5e-06, "input_cost_per_token_batches": 1e-06, + "input_cost_per_token_priority": 3.5e-06, "litellm_provider": "openai", "max_input_tokens": 1047576, "max_output_tokens": 32768, "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 8e-06, - "output_cost_per_token_priority": 1.4e-05, "output_cost_per_token_batches": 4e-06, + "output_cost_per_token_priority": 1.4e-05, "search_context_cost_per_query": { "search_context_size_high": 0.025, "search_context_size_low": 0.025, @@ -24625,16 +24658,16 @@ "cache_read_input_token_cost": 1e-07, "cache_read_input_token_cost_priority": 1.75e-07, "input_cost_per_token": 4e-07, - "input_cost_per_token_priority": 7e-07, "input_cost_per_token_batches": 2e-07, + "input_cost_per_token_priority": 7e-07, "litellm_provider": "openai", "max_input_tokens": 1047576, "max_output_tokens": 32768, "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 1.6e-06, - "output_cost_per_token_priority": 2.8e-06, "output_cost_per_token_batches": 8e-07, + "output_cost_per_token_priority": 2.8e-06, "search_context_cost_per_query": { "search_context_size_high": 0.025, "search_context_size_low": 0.025, @@ -24705,16 +24738,16 @@ "cache_read_input_token_cost_priority": 5e-08, "deprecation_date": "2026-10-23", "input_cost_per_token": 1e-07, - "input_cost_per_token_priority": 2e-07, "input_cost_per_token_batches": 5e-08, + "input_cost_per_token_priority": 2e-07, "litellm_provider": "openai", "max_input_tokens": 1047576, "max_output_tokens": 32768, "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 4e-07, - "output_cost_per_token_priority": 8e-07, "output_cost_per_token_batches": 2e-07, + "output_cost_per_token_priority": 8e-07, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -24785,16 +24818,16 @@ "cache_read_input_token_cost": 1.25e-06, "cache_read_input_token_cost_priority": 2.125e-06, "input_cost_per_token": 2.5e-06, - "input_cost_per_token_priority": 4.25e-06, "input_cost_per_token_batches": 1.25e-06, + "input_cost_per_token_priority": 4.25e-06, "litellm_provider": "openai", "max_input_tokens": 128000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", "output_cost_per_token": 1e-05, - "output_cost_per_token_priority": 1.7e-05, "output_cost_per_token_batches": 5e-06, + "output_cost_per_token_priority": 1.7e-05, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, @@ -24808,16 +24841,16 @@ "cache_read_input_token_cost": 1.25e-06, "cache_read_input_token_cost_priority": 2.125e-06, "input_cost_per_token": 2.5e-06, - "input_cost_per_token_priority": 4.25e-06, "input_cost_per_token_batches": 1.25e-06, + "input_cost_per_token_priority": 4.25e-06, "litellm_provider": "openai", "max_input_tokens": 128000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", "output_cost_per_token": 1e-05, - "output_cost_per_token_priority": 1.7e-05, "output_cost_per_token_batches": 5e-06, + "output_cost_per_token_priority": 1.7e-05, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, @@ -25063,6 +25096,7 @@ "supports_vision": false }, "gpt-audio-mini-2025-12-15": { + "deprecation_date": "2027-01-20", "input_cost_per_audio_token": 1e-05, "input_cost_per_token": 6e-07, "litellm_provider": "openai", @@ -25096,8 +25130,7 @@ "supports_response_schema": false, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": false, - "deprecation_date": "2027-01-20" + "supports_vision": false }, "gpt-4o-mini": { "cache_read_input_token_cost": 7.5e-08, @@ -25126,16 +25159,16 @@ "cache_read_input_token_cost": 7.5e-08, "cache_read_input_token_cost_priority": 1.25e-07, "input_cost_per_token": 1.5e-07, - "input_cost_per_token_priority": 2.5e-07, "input_cost_per_token_batches": 7.5e-08, + "input_cost_per_token_priority": 2.5e-07, "litellm_provider": "openai", "max_input_tokens": 128000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", "output_cost_per_token": 6e-07, - "output_cost_per_token_priority": 1e-06, "output_cost_per_token_batches": 3e-07, + "output_cost_per_token_priority": 1e-06, "search_context_cost_per_query": { "search_context_size_high": 0.03, "search_context_size_low": 0.025, @@ -25432,62 +25465,62 @@ "gpt-image-1.5": { "cache_read_input_token_cost": 1.25e-06, "deprecation_date": "2026-12-01", + "input_cost_per_image_token": 8e-06, "input_cost_per_token": 5e-06, "litellm_provider": "openai", "mode": "image_generation", - "output_cost_per_token": 1e-05, - "input_cost_per_image_token": 8e-06, "output_cost_per_image_token": 3.2e-05, + "output_cost_per_token": 1e-05, "supported_endpoints": [ "/v1/images/generations" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "gpt-image-1.5-2025-12-16": { "cache_read_input_token_cost": 1.25e-06, "deprecation_date": "2026-12-01", + "input_cost_per_image_token": 8e-06, "input_cost_per_token": 5e-06, "litellm_provider": "openai", "mode": "image_generation", - "output_cost_per_token": 1e-05, - "input_cost_per_image_token": 8e-06, "output_cost_per_image_token": 3.2e-05, + "output_cost_per_token": 1e-05, "supported_endpoints": [ "/v1/images/generations" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "gpt-image-2": { "cache_read_input_token_cost": 1.25e-06, + "input_cost_per_image_token": 8e-06, "input_cost_per_token": 5e-06, "litellm_provider": "openai", "mode": "image_generation", - "output_cost_per_token": 1e-05, - "input_cost_per_image_token": 8e-06, "output_cost_per_image_token": 3e-05, + "output_cost_per_token": 1e-05, "supported_endpoints": [ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "gpt-image-2-2026-04-21": { "cache_read_input_token_cost": 1.25e-06, + "input_cost_per_image_token": 8e-06, "input_cost_per_token": 5e-06, "litellm_provider": "openai", "mode": "image_generation", - "output_cost_per_token": 1e-05, - "input_cost_per_image_token": 8e-06, "output_cost_per_image_token": 3e-05, + "output_cost_per_token": 1e-05, "supported_endpoints": [ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "low/1024-x-1024/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25498,8 +25531,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "low/1024-x-1536/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25510,8 +25543,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "low/1536-x-1024/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25522,8 +25555,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "medium/1024-x-1024/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25534,8 +25567,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "medium/1024-x-1536/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25546,8 +25579,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "medium/1536-x-1024/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25558,8 +25591,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "high/1024-x-1024/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25570,8 +25603,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "high/1024-x-1536/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25582,8 +25615,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "high/1536-x-1024/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25594,8 +25627,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "standard/1024-x-1024/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25606,8 +25639,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "standard/1024-x-1536/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25618,8 +25651,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "standard/1536-x-1024/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25630,8 +25663,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "1024-x-1024/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25642,8 +25675,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "1024-x-1536/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25654,8 +25687,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "1536-x-1024/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25666,8 +25699,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "low/1024-x-1024/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.009, @@ -25677,8 +25710,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "low/1024-x-1536/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.013, @@ -25688,8 +25721,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "low/1536-x-1024/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.013, @@ -25699,8 +25732,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "medium/1024-x-1024/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.034, @@ -25710,8 +25743,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "medium/1024-x-1536/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.05, @@ -25721,8 +25754,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "medium/1536-x-1024/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.05, @@ -25732,8 +25765,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "high/1024-x-1024/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.133, @@ -25743,8 +25776,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "high/1024-x-1536/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.2, @@ -25754,8 +25787,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "high/1536-x-1024/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.2, @@ -25765,8 +25798,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "standard/1024-x-1024/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.009, @@ -25776,8 +25809,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "standard/1024-x-1536/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.013, @@ -25787,8 +25820,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "standard/1536-x-1024/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.013, @@ -25798,8 +25831,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "1024-x-1024/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.009, @@ -25809,8 +25842,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "1024-x-1536/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.013, @@ -25820,8 +25853,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "1536-x-1024/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.013, @@ -25831,8 +25864,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "gpt-5": { "cache_read_input_token_cost": 1.25e-07, @@ -25867,7 +25900,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -25877,9 +25912,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5.1": { "cache_read_input_token_cost": 1.25e-07, @@ -25911,7 +25944,9 @@ "image" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -25921,9 +25956,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5.1-2025-11-13": { "cache_read_input_token_cost": 1.25e-07, @@ -25955,7 +25988,9 @@ "image" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -25965,9 +26000,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5.1-chat-latest": { "cache_read_input_token_cost": 1.25e-07, @@ -26000,7 +26033,9 @@ "image" ], "supports_function_calling": false, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": false, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26010,9 +26045,7 @@ "supports_tool_choice": false, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5.2": { "cache_read_input_token_cost": 1.75e-07, @@ -26045,7 +26078,9 @@ "image" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26055,9 +26090,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "gpt-5.2-2025-12-11": { "cache_read_input_token_cost": 1.75e-07, @@ -26090,7 +26123,9 @@ "image" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26100,9 +26135,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "gpt-5.2-chat-latest": { "cache_read_input_token_cost": 1.75e-07, @@ -26134,7 +26167,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26144,9 +26179,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5.3-chat-latest": { "cache_read_input_token_cost": 1.75e-07, @@ -26178,7 +26211,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26188,9 +26223,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5.2-pro": { "input_cost_per_token": 2.1e-05, @@ -26217,7 +26250,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26227,9 +26262,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "gpt-5.2-pro-2025-12-11": { "input_cost_per_token": 2.1e-05, @@ -26256,7 +26289,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26266,9 +26301,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "gpt-5.6": { "cache_creation_input_token_cost": 5e-06, @@ -26368,6 +26401,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://developers.openai.com/api/docs/models/gpt-5.6-sol", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -26391,13 +26425,12 @@ "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_xhigh_reasoning_effort": true, - "supports_service_tier": true, - "source": "https://developers.openai.com/api/docs/models/gpt-5.6-sol" + "supports_xhigh_reasoning_effort": true }, "gpt-5.6-terra": { "cache_creation_input_token_cost": 2.5e-06, @@ -26434,6 +26467,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://developers.openai.com/api/docs/models/gpt-5.6-terra", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -26456,13 +26490,12 @@ "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_xhigh_reasoning_effort": true, - "supports_service_tier": true, - "source": "https://developers.openai.com/api/docs/models/gpt-5.6-terra" + "supports_xhigh_reasoning_effort": true }, "gpt-5.6-luna": { "cache_creation_input_token_cost": 2.5e-07, @@ -26499,6 +26532,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://developers.openai.com/api/docs/models/gpt-5.6-luna", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -26521,13 +26555,12 @@ "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_xhigh_reasoning_effort": true, - "supports_service_tier": true, - "source": "https://developers.openai.com/api/docs/models/gpt-5.6-luna" + "supports_xhigh_reasoning_effort": true }, "gpt-5.6-cyber": { "cache_creation_input_token_cost": 1.5625e-05, @@ -26543,6 +26576,7 @@ "mode": "chat", "output_cost_per_token": 7.5e-05, "output_cost_per_token_above_272k_tokens": 0.0001125, + "source": "https://platform.openai.com/docs/models/gpt-5.6-cyber", "supported_endpoints": [ "/v1/responses" ], @@ -26553,8 +26587,10 @@ "supported_output_modalities": [ "text" ], + "supports_computer_use": true, "supports_function_calling": true, "supports_native_streaming": true, + "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -26562,10 +26598,7 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "source": "https://platform.openai.com/docs/models/gpt-5.6-cyber", - "supports_computer_use": true, - "supports_parallel_function_calling": true + "supports_web_search": true }, "daybreak-red-latest": { "cache_creation_input_token_cost": 1.5625e-05, @@ -26581,6 +26614,7 @@ "mode": "chat", "output_cost_per_token": 7.5e-05, "output_cost_per_token_above_272k_tokens": 0.0001125, + "source": "https://platform.openai.com/docs/models/daybreak-red-latest", "supported_endpoints": [ "/v1/responses" ], @@ -26591,8 +26625,10 @@ "supported_output_modalities": [ "text" ], + "supports_computer_use": true, "supports_function_calling": true, "supports_native_streaming": true, + "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -26600,10 +26636,7 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "source": "https://platform.openai.com/docs/models/daybreak-red-latest", - "supports_computer_use": true, - "supports_parallel_function_calling": true + "supports_web_search": true }, "daybreak-blue-latest": { "cache_creation_input_token_cost": 5e-06, @@ -26619,6 +26652,7 @@ "mode": "chat", "output_cost_per_token": 2e-05, "output_cost_per_token_above_272k_tokens": 3e-05, + "source": "https://platform.openai.com/docs/models/daybreak-blue-latest", "supported_endpoints": [ "/v1/responses" ], @@ -26632,6 +26666,7 @@ "supports_computer_use": true, "supports_function_calling": true, "supports_native_streaming": true, + "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -26639,9 +26674,7 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "source": "https://platform.openai.com/docs/models/daybreak-blue-latest", - "supports_parallel_function_calling": true + "supports_web_search": true }, "chat-latest": { "cache_read_input_token_cost": 5e-07, @@ -26652,6 +26685,11 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 3e-05, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, "source": "https://platform.openai.com/docs/models/chat-latest", "supported_endpoints": [ "/v1/chat/completions", @@ -26664,6 +26702,7 @@ "supported_output_modalities": [ "text" ], + "supports_file_search": true, "supports_function_calling": true, "supports_native_streaming": true, "supports_parallel_function_calling": true, @@ -26673,13 +26712,7 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_high": 0.01, - "search_context_size_low": 0.01, - "search_context_size_medium": 0.01 - }, - "supports_file_search": true + "supports_web_search": true }, "gpt-5.5": { "cache_read_input_token_cost": 5e-07, @@ -26688,8 +26721,8 @@ "cache_read_input_token_cost_priority": 1e-06, "input_cost_per_token": 5e-06, "input_cost_per_token_above_272k_tokens": 1e-05, - "input_cost_per_token_flex": 2.5e-06, "input_cost_per_token_batches": 2.5e-06, + "input_cost_per_token_flex": 2.5e-06, "input_cost_per_token_priority": 1e-05, "litellm_provider": "openai", "max_input_tokens": 1050000, @@ -26698,8 +26731,8 @@ "mode": "chat", "output_cost_per_token": 3e-05, "output_cost_per_token_above_272k_tokens": 4.5e-05, - "output_cost_per_token_flex": 1.5e-05, "output_cost_per_token_batches": 1.5e-05, + "output_cost_per_token_flex": 1.5e-05, "output_cost_per_token_priority": 6e-05, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, @@ -26721,7 +26754,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26731,9 +26766,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "gpt-5.5-2026-04-23": { "cache_read_input_token_cost": 5e-07, @@ -26742,8 +26775,8 @@ "cache_read_input_token_cost_priority": 1e-06, "input_cost_per_token": 5e-06, "input_cost_per_token_above_272k_tokens": 1e-05, - "input_cost_per_token_flex": 2.5e-06, "input_cost_per_token_batches": 2.5e-06, + "input_cost_per_token_flex": 2.5e-06, "input_cost_per_token_priority": 1e-05, "litellm_provider": "openai", "max_input_tokens": 1050000, @@ -26752,8 +26785,8 @@ "mode": "chat", "output_cost_per_token": 3e-05, "output_cost_per_token_above_272k_tokens": 4.5e-05, - "output_cost_per_token_flex": 1.5e-05, "output_cost_per_token_batches": 1.5e-05, + "output_cost_per_token_flex": 1.5e-05, "output_cost_per_token_priority": 6e-05, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, @@ -26775,7 +26808,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26785,17 +26820,15 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "gpt-5.5-pro": { "cache_read_input_token_cost": 3e-06, "cache_read_input_token_cost_above_272k_tokens": 6e-06, "input_cost_per_token": 3e-05, "input_cost_per_token_above_272k_tokens": 6e-05, - "input_cost_per_token_flex": 1.5e-05, "input_cost_per_token_batches": 1.5e-05, + "input_cost_per_token_flex": 1.5e-05, "litellm_provider": "openai", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -26803,8 +26836,8 @@ "mode": "responses", "output_cost_per_token": 0.00018, "output_cost_per_token_above_272k_tokens": 0.00027, - "output_cost_per_token_flex": 9e-05, "output_cost_per_token_batches": 9e-05, + "output_cost_per_token_flex": 9e-05, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, "search_context_cost_per_query": { @@ -26824,7 +26857,10 @@ "text" ], "supports_function_calling": true, + "supports_low_reasoning_effort": false, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26834,18 +26870,15 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "supports_low_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "gpt-5.5-pro-2026-04-23": { "cache_read_input_token_cost": 3e-06, "cache_read_input_token_cost_above_272k_tokens": 6e-06, "input_cost_per_token": 3e-05, "input_cost_per_token_above_272k_tokens": 6e-05, - "input_cost_per_token_flex": 1.5e-05, "input_cost_per_token_batches": 1.5e-05, + "input_cost_per_token_flex": 1.5e-05, "litellm_provider": "openai", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -26853,8 +26886,8 @@ "mode": "responses", "output_cost_per_token": 0.00018, "output_cost_per_token_above_272k_tokens": 0.00027, - "output_cost_per_token_flex": 9e-05, "output_cost_per_token_batches": 9e-05, + "output_cost_per_token_flex": 9e-05, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, "search_context_cost_per_query": { @@ -26874,7 +26907,10 @@ "text" ], "supports_function_calling": true, + "supports_low_reasoning_effort": false, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26884,10 +26920,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "supports_low_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "gpt-5.4": { "cache_read_input_token_cost": 2.5e-07, @@ -26896,8 +26929,8 @@ "cache_read_input_token_cost_priority": 5e-07, "input_cost_per_token": 2.5e-06, "input_cost_per_token_above_272k_tokens": 5e-06, - "input_cost_per_token_flex": 1.25e-06, "input_cost_per_token_batches": 1.25e-06, + "input_cost_per_token_flex": 1.25e-06, "input_cost_per_token_priority": 5e-06, "litellm_provider": "openai", "max_input_tokens": 1050000, @@ -26906,8 +26939,8 @@ "mode": "chat", "output_cost_per_token": 1.5e-05, "output_cost_per_token_above_272k_tokens": 2.25e-05, - "output_cost_per_token_flex": 7.5e-06, "output_cost_per_token_batches": 7.5e-06, + "output_cost_per_token_flex": 7.5e-06, "output_cost_per_token_priority": 3e-05, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, @@ -26924,7 +26957,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26933,9 +26968,7 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "gpt-5.4-2026-03-05": { "cache_read_input_token_cost": 2.5e-07, @@ -26944,8 +26977,8 @@ "cache_read_input_token_cost_priority": 5e-07, "input_cost_per_token": 2.5e-06, "input_cost_per_token_above_272k_tokens": 5e-06, - "input_cost_per_token_flex": 1.25e-06, "input_cost_per_token_batches": 1.25e-06, + "input_cost_per_token_flex": 1.25e-06, "input_cost_per_token_priority": 5e-06, "litellm_provider": "openai", "max_input_tokens": 1050000, @@ -26954,8 +26987,8 @@ "mode": "chat", "output_cost_per_token": 1.5e-05, "output_cost_per_token_above_272k_tokens": 2.25e-05, - "output_cost_per_token_flex": 7.5e-06, "output_cost_per_token_batches": 7.5e-06, + "output_cost_per_token_flex": 7.5e-06, "output_cost_per_token_priority": 3e-05, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, @@ -26972,7 +27005,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26981,17 +27016,15 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "gpt-5.4-pro": { "cache_read_input_token_cost": 3e-06, "cache_read_input_token_cost_above_272k_tokens": 6e-06, "input_cost_per_token": 3e-05, "input_cost_per_token_above_272k_tokens": 6e-05, - "input_cost_per_token_flex": 1.5e-05, "input_cost_per_token_batches": 1.5e-05, + "input_cost_per_token_flex": 1.5e-05, "litellm_provider": "openai", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -26999,8 +27032,8 @@ "mode": "responses", "output_cost_per_token": 0.00018, "output_cost_per_token_above_272k_tokens": 0.00027, - "output_cost_per_token_flex": 9e-05, "output_cost_per_token_batches": 9e-05, + "output_cost_per_token_flex": 9e-05, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, "search_context_cost_per_query": { @@ -27020,7 +27053,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27030,17 +27065,15 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "gpt-5.4-pro-2026-03-05": { "cache_read_input_token_cost": 3e-06, "cache_read_input_token_cost_above_272k_tokens": 6e-06, "input_cost_per_token": 3e-05, "input_cost_per_token_above_272k_tokens": 6e-05, - "input_cost_per_token_flex": 1.5e-05, "input_cost_per_token_batches": 1.5e-05, + "input_cost_per_token_flex": 1.5e-05, "litellm_provider": "openai", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -27048,8 +27081,8 @@ "mode": "responses", "output_cost_per_token": 0.00018, "output_cost_per_token_above_272k_tokens": 0.00027, - "output_cost_per_token_flex": 9e-05, "output_cost_per_token_batches": 9e-05, + "output_cost_per_token_flex": 9e-05, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, "search_context_cost_per_query": { @@ -27069,7 +27102,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27079,17 +27114,15 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "gpt-5.4-mini": { "cache_read_input_token_cost": 7.5e-08, "cache_read_input_token_cost_flex": 3.75e-08, "cache_read_input_token_cost_priority": 1.5e-07, "input_cost_per_token": 7.5e-07, - "input_cost_per_token_flex": 3.75e-07, "input_cost_per_token_batches": 3.75e-07, + "input_cost_per_token_flex": 3.75e-07, "input_cost_per_token_priority": 1.5e-06, "litellm_provider": "openai", "max_input_tokens": 272000, @@ -27097,8 +27130,8 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 4.5e-06, - "output_cost_per_token_flex": 2.25e-06, "output_cost_per_token_batches": 2.25e-06, + "output_cost_per_token_flex": 2.25e-06, "output_cost_per_token_priority": 9e-06, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, @@ -27120,7 +27153,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27130,17 +27165,15 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "gpt-5.4-mini-2026-03-17": { "cache_read_input_token_cost": 7.5e-08, "cache_read_input_token_cost_flex": 3.75e-08, "cache_read_input_token_cost_priority": 1.5e-07, "input_cost_per_token": 7.5e-07, - "input_cost_per_token_flex": 3.75e-07, "input_cost_per_token_batches": 3.75e-07, + "input_cost_per_token_flex": 3.75e-07, "input_cost_per_token_priority": 1.5e-06, "litellm_provider": "openai", "max_input_tokens": 272000, @@ -27148,8 +27181,8 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 4.5e-06, - "output_cost_per_token_flex": 2.25e-06, "output_cost_per_token_batches": 2.25e-06, + "output_cost_per_token_flex": 2.25e-06, "output_cost_per_token_priority": 9e-06, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, @@ -27171,7 +27204,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27181,24 +27216,22 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "gpt-5.4-nano": { "cache_read_input_token_cost": 2e-08, "cache_read_input_token_cost_flex": 1e-08, "input_cost_per_token": 2e-07, - "input_cost_per_token_flex": 1e-07, "input_cost_per_token_batches": 1e-07, + "input_cost_per_token_flex": 1e-07, "litellm_provider": "openai", "max_input_tokens": 272000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1.25e-06, - "output_cost_per_token_flex": 6.25e-07, "output_cost_per_token_batches": 6.25e-07, + "output_cost_per_token_flex": 6.25e-07, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, "search_context_cost_per_query": { @@ -27219,7 +27252,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27229,24 +27264,22 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "gpt-5.4-nano-2026-03-17": { "cache_read_input_token_cost": 2e-08, "cache_read_input_token_cost_flex": 1e-08, "input_cost_per_token": 2e-07, - "input_cost_per_token_flex": 1e-07, "input_cost_per_token_batches": 1e-07, + "input_cost_per_token_flex": 1e-07, "litellm_provider": "openai", "max_input_tokens": 272000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1.25e-06, - "output_cost_per_token_flex": 6.25e-07, "output_cost_per_token_batches": 6.25e-07, + "output_cost_per_token_flex": 6.25e-07, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, "search_context_cost_per_query": { @@ -27267,7 +27300,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27277,9 +27312,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "gpt-5-pro": { "input_cost_per_token": 1.5e-05, @@ -27308,7 +27341,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": false, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27318,9 +27353,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5-pro-2025-10-06": { "deprecation_date": "2026-12-11", @@ -27350,7 +27383,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": false, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27360,9 +27395,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5-2025-08-07": { "cache_read_input_token_cost": 1.25e-07, @@ -27398,7 +27431,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27408,9 +27443,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5-chat": { "cache_read_input_token_cost": 1.25e-07, @@ -27434,7 +27467,9 @@ "text" ], "supports_function_calling": false, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": false, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27443,9 +27478,7 @@ "supports_system_messages": true, "supports_tool_choice": false, "supports_vision": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5-chat-latest": { "cache_read_input_token_cost": 1.25e-07, @@ -27470,7 +27503,9 @@ "text" ], "supports_function_calling": false, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": false, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27479,9 +27514,7 @@ "supports_system_messages": true, "supports_tool_choice": false, "supports_vision": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5-codex": { "cache_read_input_token_cost": 1.25e-07, @@ -27509,7 +27542,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27519,9 +27554,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5.1-codex": { "cache_read_input_token_cost": 1.25e-07, @@ -27552,7 +27585,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27562,9 +27597,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5.1-codex-max": { "cache_read_input_token_cost": 1.25e-07, @@ -27592,7 +27625,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27602,9 +27637,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "gpt-5.1-codex-mini": { "cache_read_input_token_cost": 2.5e-08, @@ -27635,7 +27668,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27645,9 +27680,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5.2-codex": { "cache_read_input_token_cost": 1.75e-07, @@ -27678,7 +27711,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27688,9 +27723,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "gpt-5.3-codex": { "cache_read_input_token_cost": 1.75e-07, @@ -27720,7 +27753,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27730,9 +27765,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5-mini": { "cache_read_input_token_cost": 2.5e-08, @@ -27767,7 +27800,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27777,9 +27812,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5-mini-2025-08-07": { "cache_read_input_token_cost": 2.5e-08, @@ -27815,7 +27848,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27825,9 +27860,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5-nano": { "cache_read_input_token_cost": 5e-09, @@ -27860,7 +27893,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27870,17 +27905,15 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5-nano-2025-08-07": { "cache_read_input_token_cost": 5e-09, "cache_read_input_token_cost_flex": 2.5e-09, "deprecation_date": "2026-12-11", "input_cost_per_token": 5e-08, - "input_cost_per_token_priority": 2.5e-06, "input_cost_per_token_flex": 2.5e-08, + "input_cost_per_token_priority": 2.5e-06, "litellm_provider": "openai", "max_input_tokens": 272000, "max_output_tokens": 128000, @@ -27906,7 +27939,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27916,9 +27951,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-image-1": { "cache_read_input_token_cost": 1.25e-06, @@ -28116,6 +28149,7 @@ "gpt-realtime-mini": { "cache_creation_input_audio_token_cost": 3e-07, "cache_read_input_audio_token_cost": 3e-07, + "cache_read_input_token_cost": 6e-08, "deprecation_date": "2027-01-20", "input_cost_per_audio_token": 1e-05, "input_cost_per_token": 6e-07, @@ -28143,8 +28177,7 @@ "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "cache_read_input_token_cost": 6e-08 + "supports_tool_choice": true }, "gpt-realtime-2025-08-28": { "cache_creation_input_audio_token_cost": 4e-07, @@ -28181,6 +28214,8 @@ }, "gradient_ai/alibaba-qwen3-32b": { "litellm_provider": "gradient_ai", + "max_input_tokens": 131072, + "max_output_tokens": 40960, "max_tokens": 40960, "mode": "chat", "supported_endpoints": [ @@ -28189,13 +28224,13 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 131072, - "max_output_tokens": 40960 + "supports_tool_choice": false }, "gradient_ai/anthropic-claude-3-opus": { "input_cost_per_token": 1.5e-05, "litellm_provider": "gradient_ai", + "max_input_tokens": 200000, + "max_output_tokens": 1024, "max_tokens": 1024, "mode": "chat", "output_cost_per_token": 7.5e-05, @@ -28205,13 +28240,13 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 200000, - "max_output_tokens": 1024 + "supports_tool_choice": false }, "gradient_ai/anthropic-claude-3.5-haiku": { "input_cost_per_token": 8e-07, "litellm_provider": "gradient_ai", + "max_input_tokens": 200000, + "max_output_tokens": 1024, "max_tokens": 1024, "mode": "chat", "output_cost_per_token": 4e-06, @@ -28221,13 +28256,13 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 200000, - "max_output_tokens": 1024 + "supports_tool_choice": false }, "gradient_ai/anthropic-claude-3.5-sonnet": { "input_cost_per_token": 3e-06, "litellm_provider": "gradient_ai", + "max_input_tokens": 200000, + "max_output_tokens": 1024, "max_tokens": 1024, "mode": "chat", "output_cost_per_token": 1.5e-05, @@ -28237,13 +28272,13 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 200000, - "max_output_tokens": 1024 + "supports_tool_choice": false }, "gradient_ai/anthropic-claude-3.7-sonnet": { "input_cost_per_token": 3e-06, "litellm_provider": "gradient_ai", + "max_input_tokens": 200000, + "max_output_tokens": 1024, "max_tokens": 1024, "mode": "chat", "output_cost_per_token": 1.5e-05, @@ -28253,13 +28288,13 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 200000, - "max_output_tokens": 1024 + "supports_tool_choice": false }, "gradient_ai/deepseek-r1-distill-llama-70b": { "input_cost_per_token": 9.9e-07, "litellm_provider": "gradient_ai", + "max_input_tokens": 32768, + "max_output_tokens": 8000, "max_tokens": 8000, "mode": "chat", "output_cost_per_token": 9.9e-07, @@ -28269,13 +28304,13 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 32768, - "max_output_tokens": 8000 + "supports_tool_choice": false }, "gradient_ai/llama3-8b-instruct": { "input_cost_per_token": 2e-07, "litellm_provider": "gradient_ai", + "max_input_tokens": 8192, + "max_output_tokens": 512, "max_tokens": 512, "mode": "chat", "output_cost_per_token": 2e-07, @@ -28285,13 +28320,13 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 8192, - "max_output_tokens": 512 + "supports_tool_choice": false }, "gradient_ai/llama3.3-70b-instruct": { "input_cost_per_token": 6.5e-07, "litellm_provider": "gradient_ai", + "max_input_tokens": 128000, + "max_output_tokens": 2048, "max_tokens": 2048, "mode": "chat", "output_cost_per_token": 6.5e-07, @@ -28301,13 +28336,13 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 128000, - "max_output_tokens": 2048 + "supports_tool_choice": false }, "gradient_ai/mistral-nemo-instruct-2407": { "input_cost_per_token": 3e-07, "litellm_provider": "gradient_ai", + "max_input_tokens": 128000, + "max_output_tokens": 512, "max_tokens": 512, "mode": "chat", "output_cost_per_token": 3e-07, @@ -28317,12 +28352,12 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 128000, - "max_output_tokens": 512 + "supports_tool_choice": false }, "gradient_ai/openai-gpt-4o": { "litellm_provider": "gradient_ai", + "max_input_tokens": 128000, + "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", "supported_endpoints": [ @@ -28331,12 +28366,12 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 128000, - "max_output_tokens": 16384 + "supports_tool_choice": false }, "gradient_ai/openai-gpt-4o-mini": { "litellm_provider": "gradient_ai", + "max_input_tokens": 128000, + "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", "supported_endpoints": [ @@ -28345,13 +28380,13 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 128000, - "max_output_tokens": 16384 + "supports_tool_choice": false }, "gradient_ai/openai-o3": { "input_cost_per_token": 2e-06, "litellm_provider": "gradient_ai", + "max_input_tokens": 200000, + "max_output_tokens": 100000, "max_tokens": 100000, "mode": "chat", "output_cost_per_token": 8e-06, @@ -28361,13 +28396,13 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 200000, - "max_output_tokens": 100000 + "supports_tool_choice": false }, "gradient_ai/openai-o3-mini": { "input_cost_per_token": 1.1e-06, "litellm_provider": "gradient_ai", + "max_input_tokens": 200000, + "max_output_tokens": 100000, "max_tokens": 100000, "mode": "chat", "output_cost_per_token": 4.4e-06, @@ -28377,16 +28412,14 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 200000, - "max_output_tokens": 100000 + "supports_tool_choice": false }, "lemonade/Qwen3-Coder-30B-A3B-Instruct-GGUF": { "input_cost_per_token": 0, "litellm_provider": "lemonade", - "max_tokens": 32768, "max_input_tokens": 262144, "max_output_tokens": 32768, + "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 0, "supports_function_calling": true, @@ -28396,9 +28429,9 @@ "lemonade/gpt-oss-20b-mxfp4-GGUF": { "input_cost_per_token": 0, "litellm_provider": "lemonade", - "max_tokens": 32768, "max_input_tokens": 131072, "max_output_tokens": 32768, + "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 0, "supports_function_calling": true, @@ -28408,9 +28441,9 @@ "lemonade/gpt-oss-120b-mxfp-GGUF": { "input_cost_per_token": 0, "litellm_provider": "lemonade", - "max_tokens": 32768, "max_input_tokens": 131072, "max_output_tokens": 32768, + "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 0, "supports_function_calling": true, @@ -28420,9 +28453,9 @@ "lemonade/Gemma-3-4b-it-GGUF": { "input_cost_per_token": 0, "litellm_provider": "lemonade", - "max_tokens": 8192, "max_input_tokens": 128000, "max_output_tokens": 8192, + "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 0, "supports_function_calling": true, @@ -28432,9 +28465,9 @@ "lemonade/Qwen3-4B-Instruct-2507-GGUF": { "input_cost_per_token": 0, "litellm_provider": "lemonade", - "max_tokens": 32768, "max_input_tokens": 262144, "max_output_tokens": 32768, + "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 0, "supports_function_calling": true, @@ -28592,15 +28625,15 @@ "supports_vision": true }, "groq/moonshotai/kimi-k2-instruct-0905": { + "cache_read_input_token_cost": 5e-07, "deprecation_date": "2026-04-15", "input_cost_per_token": 1e-06, - "output_cost_per_token": 3e-06, - "cache_read_input_token_cost": 5e-07, "litellm_provider": "groq", "max_input_tokens": 262144, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 3e-06, "supports_function_calling": true, "supports_response_schema": true, "supports_tool_choice": true @@ -28695,6 +28728,7 @@ "mode": "audio_speech" }, "groq/qwen/qwen3.6-27b": { + "cache_read_input_token_cost": 3e-07, "input_cost_per_token": 6e-07, "litellm_provider": "groq", "max_input_tokens": 131072, @@ -28707,8 +28741,7 @@ "supports_reasoning": true, "supports_response_schema": false, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-07 + "supports_vision": true }, "groq/qwen/qwen3-32b": { "deprecation_date": "2026-07-17", @@ -28759,43 +28792,43 @@ }, "heroku/claude-3-5-haiku": { "litellm_provider": "heroku", + "max_input_tokens": 200000, + "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "max_input_tokens": 200000, - "max_output_tokens": 8192 + "supports_tool_choice": true }, "heroku/claude-3-5-sonnet-latest": { "litellm_provider": "heroku", + "max_input_tokens": 200000, + "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "max_input_tokens": 200000, - "max_output_tokens": 8192 + "supports_tool_choice": true }, "heroku/claude-3-7-sonnet": { "litellm_provider": "heroku", + "max_input_tokens": 200000, + "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "max_input_tokens": 200000, - "max_output_tokens": 8192 + "supports_tool_choice": true }, "heroku/claude-4-sonnet": { "litellm_provider": "heroku", + "max_input_tokens": 200000, + "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "max_input_tokens": 200000, - "max_output_tokens": 8192 + "supports_tool_choice": true }, "high/1024-x-1024/gpt-image-1": { "deprecation_date": "2026-10-23", @@ -29167,20 +29200,22 @@ "jp.anthropic.claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 4.125e-06, "cache_creation_input_token_cost_above_1hr": 6.6e-06, + "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.32e-05, + "cache_creation_input_token_cost_above_200k_tokens": 8.25e-06, "cache_read_input_token_cost": 3.3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6.6e-07, "input_cost_per_token": 3.3e-06, "input_cost_per_token_above_200k_tokens": 6.6e-06, - "output_cost_per_token_above_200k_tokens": 2.475e-05, - "cache_creation_input_token_cost_above_200k_tokens": 8.25e-06, - "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.32e-05, - "cache_read_input_token_cost_above_200k_tokens": 6.6e-07, + "input_cost_per_token_batches": 1.65e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.65e-05, + "output_cost_per_token_above_200k_tokens": 2.475e-05, + "output_cost_per_token_batches": 8.25e-06, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -29189,45 +29224,43 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024, - "input_cost_per_token_batches": 1.65e-06, - "output_cost_per_token_batches": 8.25e-06 + "supports_tool_search": true, + "supports_vision": true }, "jp.anthropic.claude-haiku-4-5-20251001-v1:0": { "cache_creation_input_token_cost": 1.375e-06, "cache_creation_input_token_cost_above_1hr": 2.2e-06, "cache_read_input_token_cost": 1.1e-07, "input_cost_per_token": 1.1e-06, + "input_cost_per_token_batches": 5.5e-07, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5.5e-06, + "output_cost_per_token_batches": 2.75e-06, + "prompt_cache_min_tokens": 4096, "source": "https://aws.amazon.com/about-aws/whats-new/2025/10/claude-4-5-haiku-anthropic-amazon-bedrock", "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096, - "input_cost_per_token_batches": 5.5e-07, - "output_cost_per_token_batches": 2.75e-06 + "supports_tool_search": true, + "supports_vision": true }, "crusoe/deepseek-ai/DeepSeek-R1-0528": { "input_cost_per_token": 3e-06, @@ -30132,8 +30165,8 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.2e-06, - "supports_system_messages": true, - "supports_native_structured_output": true + "supports_native_structured_output": true, + "supports_system_messages": true }, "minimax.minimax-m2.1": { "input_cost_per_token": 3e-07, @@ -30143,23 +30176,23 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "minimax.minimax-m2.5": { "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "minimax/speech-02-hd": { "input_cost_per_character": 0.0001, @@ -30194,98 +30227,98 @@ ] }, "minimax/MiniMax-M2.1": { - "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, - "cache_read_input_token_cost": 3e-08, "cache_creation_input_token_cost": 3.75e-07, + "cache_read_input_token_cost": 3e-08, + "input_cost_per_token": 3e-07, "litellm_provider": "minimax", + "max_input_tokens": 1000000, + "max_output_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.2e-06, "supports_function_calling": true, - "supports_tool_choice": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_system_messages": true, - "max_input_tokens": 1000000, - "max_output_tokens": 8192 + "supports_tool_choice": true }, "minimax/MiniMax-M2.1-lightning": { - "input_cost_per_token": 3e-07, - "output_cost_per_token": 2.4e-06, - "cache_read_input_token_cost": 3e-08, "cache_creation_input_token_cost": 3.75e-07, + "cache_read_input_token_cost": 3e-08, + "input_cost_per_token": 3e-07, "litellm_provider": "minimax", + "max_input_tokens": 1000000, + "max_output_tokens": 8192, "mode": "chat", + "output_cost_per_token": 2.4e-06, "supports_function_calling": true, - "supports_tool_choice": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_system_messages": true, - "max_input_tokens": 1000000, - "max_output_tokens": 8192 + "supports_tool_choice": true }, "minimax/MiniMax-M2.5": { - "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, - "cache_read_input_token_cost": 3e-08, "cache_creation_input_token_cost": 3.75e-07, + "cache_read_input_token_cost": 3e-08, + "input_cost_per_token": 3e-07, "litellm_provider": "minimax", + "max_input_tokens": 1000000, + "max_output_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.2e-06, "supports_function_calling": true, - "supports_tool_choice": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_system_messages": true, - "max_input_tokens": 1000000, - "max_output_tokens": 8192 + "supports_tool_choice": true }, "minimax/MiniMax-M2.5-lightning": { - "input_cost_per_token": 3e-07, - "output_cost_per_token": 2.4e-06, - "cache_read_input_token_cost": 3e-08, "cache_creation_input_token_cost": 3.75e-07, + "cache_read_input_token_cost": 3e-08, + "input_cost_per_token": 3e-07, "litellm_provider": "minimax", + "max_input_tokens": 1000000, + "max_output_tokens": 8192, "mode": "chat", + "output_cost_per_token": 2.4e-06, "supports_function_calling": true, - "supports_tool_choice": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_system_messages": true, - "max_input_tokens": 1000000, - "max_output_tokens": 8192 + "supports_tool_choice": true }, "minimax/MiniMax-M2": { - "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, - "cache_read_input_token_cost": 3e-08, "cache_creation_input_token_cost": 3.75e-07, + "cache_read_input_token_cost": 3e-08, + "input_cost_per_token": 3e-07, "litellm_provider": "minimax", + "max_input_tokens": 200000, + "max_output_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.2e-06, "supports_function_calling": true, - "supports_tool_choice": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_system_messages": true, - "max_input_tokens": 200000, - "max_output_tokens": 8192 + "supports_tool_choice": true }, "minimax/MiniMax-M3": { - "input_cost_per_token": 3e-07, - "input_cost_per_token_above_512k_tokens": 6e-07, - "output_cost_per_token": 1.2e-06, - "output_cost_per_token_above_512k_tokens": 2.4e-06, "cache_read_input_token_cost": 6e-08, "cache_read_input_token_cost_above_512k_tokens": 1.2e-07, + "input_cost_per_token": 3e-07, + "input_cost_per_token_above_512k_tokens": 6e-07, "litellm_provider": "minimax", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1.2e-06, + "output_cost_per_token_above_512k_tokens": 2.4e-06, "supports_function_calling": true, - "supports_tool_choice": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_vision": true, - "max_input_tokens": 1000000, - "max_output_tokens": 128000, - "supports_video_input": true + "supports_tool_choice": true, + "supports_video_input": true, + "supports_vision": true }, "mistral.devstral-2-123b": { "input_cost_per_token": 4e-07, @@ -30295,10 +30328,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "mistral.magistral-small-2509": { "input_cost_per_token": 5e-07, @@ -30321,8 +30354,8 @@ "mode": "chat", "output_cost_per_token": 2e-07, "supports_function_calling": true, - "supports_system_messages": true, - "supports_native_structured_output": true + "supports_native_structured_output": true, + "supports_system_messages": true }, "mistral.ministral-3-3b-instruct": { "input_cost_per_token": 1e-07, @@ -30333,8 +30366,8 @@ "mode": "chat", "output_cost_per_token": 1e-07, "supports_function_calling": true, - "supports_system_messages": true, - "supports_native_structured_output": true + "supports_native_structured_output": true, + "supports_system_messages": true }, "mistral.ministral-3-8b-instruct": { "input_cost_per_token": 1.5e-07, @@ -30345,8 +30378,8 @@ "mode": "chat", "output_cost_per_token": 1.5e-07, "supports_function_calling": true, - "supports_system_messages": true, - "supports_native_structured_output": true + "supports_native_structured_output": true, + "supports_system_messages": true }, "mistral.mistral-7b-instruct-v0:2": { "input_cost_per_token": 1.5e-07, @@ -30388,8 +30421,8 @@ "mode": "chat", "output_cost_per_token": 1.5e-06, "supports_function_calling": true, - "supports_system_messages": true, - "supports_native_structured_output": true + "supports_native_structured_output": true, + "supports_system_messages": true }, "mistral.mistral-small-2402-v1:0": { "input_cost_per_token": 1e-06, @@ -30420,8 +30453,8 @@ "mode": "chat", "output_cost_per_token": 4e-08, "supports_audio_input": true, - "supports_system_messages": true, - "supports_native_structured_output": true + "supports_native_structured_output": true, + "supports_system_messages": true }, "mistral.voxtral-small-24b-2507": { "input_cost_per_token": 1e-07, @@ -30432,8 +30465,8 @@ "mode": "chat", "output_cost_per_token": 3e-07, "supports_audio_input": true, - "supports_system_messages": true, - "supports_native_structured_output": true + "supports_native_structured_output": true, + "supports_system_messages": true }, "mistral/codestral-2405": { "deprecation_date": "2025-06-16", @@ -30470,11 +30503,11 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 9e-07, - "supports_assistant_prefill": true, - "supports_response_schema": true, - "supports_tool_choice": true, "source": "https://docs.mistral.ai/models/model-cards/codestral-25-08", - "supports_function_calling": true + "supports_assistant_prefill": true, + "supports_function_calling": true, + "supports_response_schema": true, + "supports_tool_choice": true }, "mistral/codestral-mamba-latest": { "input_cost_per_token": 2.5e-07, @@ -30688,24 +30721,24 @@ "supports_tool_choice": true }, "mistral/mistral-ocr-latest": { - "litellm_provider": "mistral", - "ocr_cost_per_page": 0.004, "annotation_cost_per_page": 0.005, + "litellm_provider": "mistral", "mode": "ocr", + "ocr_cost_per_page": 0.004, + "source": "https://mistral.ai/pricing#api-pricing", "supported_endpoints": [ "/v1/ocr" - ], - "source": "https://mistral.ai/pricing#api-pricing" + ] }, "mistral/mistral-ocr-4-0": { - "litellm_provider": "mistral", - "ocr_cost_per_page": 0.004, "annotation_cost_per_page": 0.005, + "litellm_provider": "mistral", "mode": "ocr", + "ocr_cost_per_page": 0.004, + "source": "https://mistral.ai/pricing#api-pricing", "supported_endpoints": [ "/v1/ocr" - ], - "source": "https://mistral.ai/pricing#api-pricing" + ] }, "mistral/mistral-ocr-4-1": { "annotation_cost_per_page": 0.005, @@ -30718,25 +30751,25 @@ ] }, "mistral/mistral-ocr-2505-completion": { + "annotation_cost_per_page": 0.003, "deprecation_date": "2026-05-31", "litellm_provider": "mistral", - "ocr_cost_per_page": 0.001, - "annotation_cost_per_page": 0.003, "mode": "ocr", + "ocr_cost_per_page": 0.001, + "source": "https://mistral.ai/pricing#api-pricing", "supported_endpoints": [ "/v1/ocr" - ], - "source": "https://mistral.ai/pricing#api-pricing" + ] }, "mistral/mistral-ocr-2512": { - "litellm_provider": "mistral", - "ocr_cost_per_page": 0.002, "annotation_cost_per_page": 0.003, + "litellm_provider": "mistral", "mode": "ocr", + "ocr_cost_per_page": 0.002, + "source": "https://mistral.ai/pricing#api-pricing", "supported_endpoints": [ "/v1/ocr" - ], - "source": "https://mistral.ai/pricing#api-pricing" + ] }, "mistral/magistral-medium-latest": { "input_cost_per_token": 2e-06, @@ -31051,9 +31084,9 @@ "source": "https://docs.mistral.ai/models/model-cards/mistral-small-4-0-26-03", "supports_assistant_prefill": true, "supports_function_calling": true, + "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_reasoning": true, "supports_vision": true }, "mistral/mistral-small-3-2-2506": { @@ -31292,9 +31325,9 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 2.5e-06, + "supports_native_structured_output": true, "supports_reasoning": true, - "supports_system_messages": true, - "supports_native_structured_output": true + "supports_system_messages": true }, "moonshotai.kimi-k2.5": { "input_cost_per_token": 6e-07, @@ -31304,11 +31337,11 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 3e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_vision": true }, "moonshot/kimi-k2-0711-preview": { "cache_read_input_token_cost": 1.5e-07, @@ -31391,42 +31424,6 @@ "supports_video_input": true, "supports_vision": true }, - "moonshot/kimi-k2.7-code": { - "cache_read_input_token_cost": 1.9e-07, - "input_cost_per_token": 9.5e-07, - "litellm_provider": "moonshot", - "max_input_tokens": 262144, - "max_output_tokens": 262144, - "max_tokens": 262144, - "mode": "chat", - "output_cost_per_token": 4e-06, - "source": "https://platform.kimi.ai/docs/pricing/chat-k27-code", - "supports_function_calling": true, - "supports_prompt_caching": true, - "supports_reasoning": true, - "supports_response_schema": true, - "supports_tool_choice": true, - "supports_video_input": true, - "supports_vision": true - }, - "moonshot/kimi-k2.7-code-highspeed": { - "cache_read_input_token_cost": 3.8e-07, - "input_cost_per_token": 1.9e-06, - "litellm_provider": "moonshot", - "max_input_tokens": 262144, - "max_output_tokens": 262144, - "max_tokens": 262144, - "mode": "chat", - "output_cost_per_token": 8e-06, - "source": "https://platform.kimi.ai/docs/pricing/chat-k27-code", - "supports_function_calling": true, - "supports_prompt_caching": true, - "supports_reasoning": true, - "supports_response_schema": true, - "supports_tool_choice": true, - "supports_video_input": true, - "supports_vision": true - }, "moonshot/kimi-k3": { "cache_read_input_token_cost": 3e-07, "input_cost_per_token": 3e-06, @@ -31712,6 +31709,7 @@ "supports_vision": false }, "multimodalembedding": { + "deprecation_date": "2027-04-01", "input_cost_per_character": 2e-07, "input_cost_per_image": 0.0001, "input_cost_per_token": 8e-07, @@ -31732,8 +31730,7 @@ "text", "image", "video" - ], - "deprecation_date": "2027-04-01" + ] }, "multimodalembedding@001": { "deprecation_date": "2027-04-01", @@ -31905,332 +31902,332 @@ ] }, "nebius/deepseek-ai/DeepSeek-R1": { - "max_tokens": 128000, + "input_cost_per_token": 8e-07, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 8e-07, - "output_cost_per_token": 2.4e-06, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 2.4e-06, + "source": "https://nebius.com/prices-ai-studio", "supports_function_calling": true, - "supports_reasoning": true, - "source": "https://nebius.com/prices-ai-studio" + "supports_reasoning": true }, "nebius/deepseek-ai/DeepSeek-R1-0528": { - "max_tokens": 164000, + "input_cost_per_token": 8e-07, + "litellm_provider": "nebius", "max_input_tokens": 164000, "max_output_tokens": 164000, - "input_cost_per_token": 8e-07, - "output_cost_per_token": 2.4e-06, - "litellm_provider": "nebius", + "max_tokens": 164000, "mode": "chat", + "output_cost_per_token": 2.4e-06, + "source": "https://nebius.com/prices-ai-studio", "supports_function_calling": true, - "supports_reasoning": true, - "source": "https://nebius.com/prices-ai-studio" + "supports_reasoning": true }, "nebius/deepseek-ai/DeepSeek-R1-Distill-Llama-70B": { - "max_tokens": 128000, + "input_cost_per_token": 2.5e-07, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 2.5e-07, - "output_cost_per_token": 7.5e-07, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 7.5e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/deepseek-ai/DeepSeek-V3": { - "max_tokens": 128000, + "input_cost_per_token": 5e-07, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 1.5e-06, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 1.5e-06, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/deepseek-ai/DeepSeek-V3-0324": { - "max_tokens": 128000, + "input_cost_per_token": 5e-07, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 1.5e-06, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 1.5e-06, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/google/gemma-3-27b-it": { - "max_tokens": 128000, + "input_cost_per_token": 6e-08, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 6e-08, - "output_cost_per_token": 2e-07, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 2e-07, + "source": "https://nebius.com/prices-ai-studio", "supports_function_calling": true, - "supports_vision": true, - "source": "https://nebius.com/prices-ai-studio" + "supports_vision": true }, "nebius/meta-llama/Llama-3.3-70B-Instruct": { - "max_tokens": 128000, + "input_cost_per_token": 1.3e-07, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 1.3e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 4e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/meta-llama/Llama-Guard-3-8B": { - "max_tokens": 128000, + "input_cost_per_token": 2e-08, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 2e-08, - "output_cost_per_token": 6e-08, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 6e-08, "source": "https://nebius.com/prices-ai-studio" }, "nebius/meta-llama/Meta-Llama-3.1-8B-Instruct": { - "max_tokens": 128000, + "input_cost_per_token": 2e-08, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 2e-08, - "output_cost_per_token": 6e-08, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 6e-08, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/meta-llama/Meta-Llama-3.1-70B-Instruct": { - "max_tokens": 128000, + "input_cost_per_token": 1.3e-07, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 1.3e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 4e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/meta-llama/Meta-Llama-3.1-405B-Instruct": { - "max_tokens": 128000, + "input_cost_per_token": 1e-06, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 1e-06, - "output_cost_per_token": 3e-06, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 3e-06, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/mistralai/Mistral-Nemo-Instruct-2407": { - "max_tokens": 128000, + "input_cost_per_token": 4e-08, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 4e-08, - "output_cost_per_token": 1.2e-07, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 1.2e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/NousResearch/Hermes-3-Llama-3.1-405B": { - "max_tokens": 128000, + "input_cost_per_token": 1e-06, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 1e-06, - "output_cost_per_token": 3e-06, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 3e-06, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/nvidia/Llama-3.1-Nemotron-Ultra-253B-v1": { - "max_tokens": 128000, + "input_cost_per_token": 6e-07, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 6e-07, - "output_cost_per_token": 1.8e-06, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 1.8e-06, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/nvidia/Llama-3.3-Nemotron-Super-49B-v1": { - "max_tokens": 131072, + "input_cost_per_token": 1e-07, + "litellm_provider": "nebius", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "nebius", + "max_tokens": 131072, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 4e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/Qwen/Qwen3-235B-A22B": { - "max_tokens": 262144, + "input_cost_per_token": 2e-07, + "litellm_provider": "nebius", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "nebius", + "max_tokens": 262144, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 6e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/Qwen/Qwen3-32B": { - "max_tokens": 32768, + "input_cost_per_token": 1e-07, + "litellm_provider": "nebius", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 3e-07, - "litellm_provider": "nebius", + "max_tokens": 32768, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 3e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/Qwen/Qwen3-30B-A3B": { - "max_tokens": 32768, + "input_cost_per_token": 1e-07, + "litellm_provider": "nebius", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 3e-07, - "litellm_provider": "nebius", + "max_tokens": 32768, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 3e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/Qwen/Qwen3-14B": { - "max_tokens": 32768, + "input_cost_per_token": 8e-08, + "litellm_provider": "nebius", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 8e-08, - "output_cost_per_token": 2.4e-07, - "litellm_provider": "nebius", + "max_tokens": 32768, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 2.4e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/Qwen/Qwen3-4B": { - "max_tokens": 32768, + "input_cost_per_token": 8e-08, + "litellm_provider": "nebius", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 8e-08, - "output_cost_per_token": 2.4e-07, - "litellm_provider": "nebius", + "max_tokens": 32768, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 2.4e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/Qwen/QwQ-32B": { - "max_tokens": 32768, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "nebius", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 4.5e-07, - "litellm_provider": "nebius", + "max_tokens": 32768, "mode": "chat", + "output_cost_per_token": 4.5e-07, + "source": "https://nebius.com/prices-ai-studio", "supports_function_calling": true, - "supports_reasoning": true, - "source": "https://nebius.com/prices-ai-studio" + "supports_reasoning": true }, "nebius/Qwen/Qwen2.5-72B-Instruct": { - "max_tokens": 128000, + "input_cost_per_token": 1.3e-07, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 1.3e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 4e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/Qwen/Qwen2.5-32B-Instruct": { - "max_tokens": 128000, + "input_cost_per_token": 6e-08, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 6e-08, - "output_cost_per_token": 2e-07, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 2e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/Qwen/Qwen2.5-Coder-7B": { - "max_tokens": 32768, + "input_cost_per_token": 1e-08, + "litellm_provider": "nebius", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-08, - "output_cost_per_token": 3e-08, - "litellm_provider": "nebius", + "max_tokens": 32768, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 3e-08, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/Qwen/Qwen2.5-VL-72B-Instruct": { - "max_tokens": 131072, + "input_cost_per_token": 1.3e-07, + "litellm_provider": "nebius", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1.3e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "nebius", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 4e-07, + "source": "https://nebius.com/prices-ai-studio", "supports_function_calling": true, - "supports_vision": true, - "source": "https://nebius.com/prices-ai-studio" + "supports_vision": true }, "nebius/Qwen/Qwen2-VL-72B-Instruct": { - "max_tokens": 131072, + "input_cost_per_token": 1.3e-07, + "litellm_provider": "nebius", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1.3e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "nebius", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 4e-07, + "source": "https://nebius.com/prices-ai-studio", "supports_function_calling": true, - "supports_vision": true, - "source": "https://nebius.com/prices-ai-studio" + "supports_vision": true }, "nebius/Qwen/Qwen2-VL-7B-Instruct": { - "max_tokens": 131072, + "input_cost_per_token": 2e-08, + "litellm_provider": "nebius", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-08, - "output_cost_per_token": 6e-08, - "litellm_provider": "nebius", + "max_tokens": 131072, "mode": "chat", - "supports_vision": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 6e-08, + "source": "https://nebius.com/prices-ai-studio", + "supports_vision": true }, "nebius/BAAI/bge-en-icl": { - "max_tokens": 32768, - "max_input_tokens": 32768, "input_cost_per_token": 1e-08, - "output_cost_per_token": 0.0, "litellm_provider": "nebius", + "max_input_tokens": 32768, + "max_tokens": 32768, "mode": "embedding", + "output_cost_per_token": 0.0, "source": "https://nebius.com/prices-ai-studio" }, "nebius/BAAI/bge-multilingual-gemma2": { - "max_tokens": 8192, - "max_input_tokens": 8192, "input_cost_per_token": 1e-08, - "output_cost_per_token": 0.0, "litellm_provider": "nebius", + "max_input_tokens": 8192, + "max_tokens": 8192, "mode": "embedding", + "output_cost_per_token": 0.0, "source": "https://nebius.com/prices-ai-studio" }, "nebius/intfloat/e5-mistral-7b-instruct": { - "max_tokens": 32768, - "max_input_tokens": 32768, "input_cost_per_token": 1e-08, - "output_cost_per_token": 0.0, "litellm_provider": "nebius", + "max_input_tokens": 32768, + "max_tokens": 32768, "mode": "embedding", + "output_cost_per_token": 0.0, "source": "https://nebius.com/prices-ai-studio" }, "nvidia.nemotron-nano-12b-v2": { @@ -32262,11 +32259,11 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 2.4e-07, - "supports_function_calling": true, - "supports_system_messages": true, - "supports_tool_choice": true, "source": "https://aws.amazon.com/bedrock/pricing/", - "supports_native_structured_output": true + "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "nvidia.nemotron-super-3-120b": { "input_cost_per_token": 1.5e-07, @@ -32816,8 +32813,8 @@ "output_cost_per_token": 7.2e-07, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/meta.llama-3.1-70b-instruct": { "input_cost_per_token": 7.2e-07, @@ -32829,8 +32826,8 @@ "output_cost_per_token": 7.2e-07, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/meta.llama-3.1-405b-instruct": { "input_cost_per_token": 1.068e-05, @@ -32842,8 +32839,8 @@ "output_cost_per_token": 1.068e-05, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/meta.llama-3.2-90b-vision-instruct": { "input_cost_per_token": 2e-06, @@ -32855,8 +32852,8 @@ "output_cost_per_token": 2e-06, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, "supports_native_streaming": true, + "supports_response_schema": false, "supports_vision": true }, "oci/meta.llama-3.3-70b-instruct": { @@ -32869,8 +32866,8 @@ "output_cost_per_token": 7.2e-07, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/meta.llama-4-maverick-17b-128e-instruct-fp8": { "input_cost_per_token": 7.2e-07, @@ -32882,8 +32879,8 @@ "output_cost_per_token": 7.2e-07, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, "supports_native_streaming": true, + "supports_response_schema": false, "supports_vision": true }, "oci/meta.llama-4-scout-17b-16e-instruct": { @@ -32896,8 +32893,8 @@ "output_cost_per_token": 7.2e-07, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/xai.grok-3": { "input_cost_per_token": 3e-06, @@ -32909,8 +32906,8 @@ "output_cost_per_token": 1.5e-05, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/xai.grok-3-fast": { "input_cost_per_token": 5e-06, @@ -32922,8 +32919,8 @@ "output_cost_per_token": 2.5e-05, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/xai.grok-3-mini": { "input_cost_per_token": 3e-07, @@ -32935,8 +32932,8 @@ "output_cost_per_token": 5e-07, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/xai.grok-3-mini-fast": { "input_cost_per_token": 6e-07, @@ -32948,8 +32945,8 @@ "output_cost_per_token": 4e-06, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/xai.grok-4": { "input_cost_per_token": 3e-06, @@ -32961,8 +32958,8 @@ "output_cost_per_token": 1.5e-05, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/cohere.command-latest": { "input_cost_per_token": 1.56e-06, @@ -32974,8 +32971,8 @@ "output_cost_per_token": 1.56e-06, "source": "https://www.oracle.com/cloud/ai/generative-ai/pricing/", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/cohere.command-a-03-2025": { "input_cost_per_token": 1.56e-06, @@ -32987,8 +32984,8 @@ "output_cost_per_token": 1.56e-06, "source": "https://www.oracle.com/cloud/ai/generative-ai/pricing/", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/cohere.command-plus-latest": { "input_cost_per_token": 1.56e-06, @@ -33000,8 +32997,8 @@ "output_cost_per_token": 1.56e-06, "source": "https://www.oracle.com/cloud/ai/generative-ai/pricing/", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/google.gemini-2.5-flash": { "input_cost_per_token": 1.5e-07, @@ -33013,10 +33010,10 @@ "output_cost_per_token": 6e-07, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": true, - "supports_vision": true, + "supports_image_size": false, "supports_native_streaming": true, - "supports_image_size": false + "supports_response_schema": true, + "supports_vision": true }, "oci/google.gemini-2.5-pro": { "input_cost_per_token": 1.25e-06, @@ -33028,9 +33025,9 @@ "output_cost_per_token": 1e-05, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, + "supports_native_streaming": true, "supports_response_schema": true, - "supports_vision": true, - "supports_native_streaming": true + "supports_vision": true }, "oci/google.gemini-2.5-flash-lite": { "input_cost_per_token": 7.5e-08, @@ -33042,10 +33039,10 @@ "output_cost_per_token": 3e-07, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": true, - "supports_vision": true, + "supports_image_size": false, "supports_native_streaming": true, - "supports_image_size": false + "supports_response_schema": true, + "supports_vision": true }, "oci/cohere.command-a-vision": { "input_cost_per_token": 1.56e-06, @@ -33057,8 +33054,8 @@ "output_cost_per_token": 1.56e-06, "source": "https://www.oracle.com/cloud/ai/generative-ai/pricing/", "supports_function_calling": true, - "supports_response_schema": false, "supports_native_streaming": true, + "supports_response_schema": false, "supports_vision": true }, "oci/cohere.command-a-reasoning": { @@ -33071,8 +33068,8 @@ "output_cost_per_token": 1.56e-06, "source": "https://www.oracle.com/cloud/ai/generative-ai/pricing/", "supports_function_calling": false, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/cohere.embed-multilingual-image-v3.0": { "input_cost_per_token": 1e-07, @@ -33704,14 +33701,14 @@ "input_cost_per_image": 0.0004, "input_cost_per_token": 2.5e-07, "litellm_provider": "openrouter", + "max_input_tokens": 200000, + "max_output_tokens": 4096, "max_tokens": 4096, "mode": "chat", "output_cost_per_token": 1.25e-06, "supports_function_calling": true, "supports_tool_choice": true, - "supports_vision": true, - "max_input_tokens": 200000, - "max_output_tokens": 4096 + "supports_vision": true }, "openrouter/anthropic/claude-3.5-sonnet": { "input_cost_per_token": 3e-06, @@ -33744,9 +33741,9 @@ "supports_vision": true }, "openrouter/anthropic/claude-opus-4": { - "input_cost_per_image": 0.0048, "cache_creation_input_token_cost": 1.875e-05, "cache_read_input_token_cost": 1.5e-06, + "input_cost_per_image": 0.0048, "input_cost_per_token": 1.5e-05, "litellm_provider": "openrouter", "max_input_tokens": 200000, @@ -33763,10 +33760,10 @@ "supports_vision": true }, "openrouter/anthropic/claude-opus-4.1": { - "input_cost_per_image": 0.0048, "cache_creation_input_token_cost": 1.875e-05, "cache_creation_input_token_cost_above_1hr": 3e-05, "cache_read_input_token_cost": 1.5e-06, + "input_cost_per_image": 0.0048, "input_cost_per_token": 1.5e-05, "litellm_provider": "openrouter", "max_input_tokens": 200000, @@ -33774,41 +33771,40 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "openrouter/anthropic/claude-sonnet-4": { - "input_cost_per_image": 0.0048, "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "input_cost_per_image": 0.0048, "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, "litellm_provider": "openrouter", "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "openrouter/anthropic/claude-sonnet-4.6": { - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, @@ -33822,16 +33818,18 @@ "mode": "chat", "output_cost_per_token": 1.5e-05, "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "source": "https://openrouter.ai/anthropic/claude-sonnet-4.6", + "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "openrouter/anthropic/claude-opus-4.5": { "cache_creation_input_token_cost": 6.25e-06, @@ -33843,18 +33841,17 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_output_config": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "openrouter/anthropic/claude-opus-4.6": { - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, @@ -33864,39 +33861,41 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, + "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "openrouter/anthropic/claude-sonnet-4.5": { - "input_cost_per_image": 0.0048, "cache_creation_input_token_cost": 3.75e-06, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "input_cost_per_image": 0.0048, "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "litellm_provider": "openrouter", "max_input_tokens": 1000000, "max_output_tokens": 1000000, "max_tokens": 1000000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "openrouter/anthropic/claude-haiku-4.5": { "cache_creation_input_token_cost": 1.25e-06, @@ -33908,17 +33907,16 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, + "prompt_cache_min_tokens": 4096, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "openrouter/anthropic/claude-opus-4.7": { - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, @@ -33928,22 +33926,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 2048, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "openrouter/anthropic/claude-opus-5": { - "prompt_cache_min_tokens": 512, - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, @@ -33953,15 +33950,17 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 512, "source": "https://openrouter.ai/anthropic/claude-opus-5", + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true @@ -34084,12 +34083,12 @@ "mode": "chat", "output_cost_per_token": 3.96e-06, "source": "https://openrouter.ai/deepseek/deepseek-v4-pro", + "supports_assistant_prefill": true, "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_tool_choice": true, - "supports_assistant_prefill": true + "supports_tool_choice": true }, "openrouter/deepseek/deepseek-v4-pro-0813": { "input_cost_per_token": 1.32e-06, @@ -34135,11 +34134,11 @@ "output_cost_per_token": 2.5e-06, "supports_audio_output": true, "supports_function_calling": true, + "supports_image_size": false, "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_image_size": false + "supports_vision": true }, "openrouter/google/gemini-2.5-pro": { "input_cost_per_audio_token": 7e-07, @@ -34158,9 +34157,9 @@ "supports_vision": true }, "openrouter/google/gemini-3-pro-preview": { + "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "input_cost_per_token": 2e-06, "input_cost_per_token_above_200k_tokens": 4e-06, "input_cost_per_token_batches": 1e-06, @@ -34326,9 +34325,9 @@ "tpm": 800000 }, "openrouter/google/gemini-3.1-pro-preview": { + "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "input_cost_per_token": 2e-06, "input_cost_per_token_above_200k_tokens": 4e-06, "litellm_provider": "openrouter", @@ -34369,22 +34368,22 @@ "openrouter/mancer/weaver": { "input_cost_per_token": 5.625e-06, "litellm_provider": "openrouter", + "max_input_tokens": 8000, + "max_output_tokens": 2000, "max_tokens": 2000, "mode": "chat", "output_cost_per_token": 5.625e-06, - "supports_tool_choice": true, - "max_input_tokens": 8000, - "max_output_tokens": 2000 + "supports_tool_choice": true }, "openrouter/meta-llama/llama-3-70b-instruct": { "input_cost_per_token": 5.9e-07, "litellm_provider": "openrouter", + "max_input_tokens": 8192, + "max_output_tokens": 8000, "max_tokens": 8000, "mode": "chat", "output_cost_per_token": 7.9e-07, - "supports_tool_choice": true, - "max_input_tokens": 8192, - "max_output_tokens": 8000 + "supports_tool_choice": true }, "openrouter/minimax/minimax-m2": { "input_cost_per_token": 2.55e-07, @@ -34472,52 +34471,52 @@ "openrouter/mistralai/mistral-7b-instruct": { "input_cost_per_token": 1.3e-07, "litellm_provider": "openrouter", + "max_input_tokens": 32768, + "max_output_tokens": 8191, "max_tokens": 8191, "mode": "chat", "output_cost_per_token": 1.3e-07, - "supports_tool_choice": true, - "max_input_tokens": 32768, - "max_output_tokens": 8191 + "supports_tool_choice": true }, "openrouter/mistralai/mistral-large": { "input_cost_per_token": 8e-06, "litellm_provider": "openrouter", + "max_input_tokens": 128000, + "max_output_tokens": 8191, "max_tokens": 8191, "mode": "chat", "output_cost_per_token": 2.4e-05, - "supports_tool_choice": true, - "max_input_tokens": 128000, - "max_output_tokens": 8191 + "supports_tool_choice": true }, "openrouter/mistralai/mistral-small-3.1-24b-instruct": { "input_cost_per_token": 1e-07, "litellm_provider": "openrouter", + "max_input_tokens": 131072, + "max_output_tokens": 131072, "max_tokens": 131072, "mode": "chat", "output_cost_per_token": 3e-07, - "supports_tool_choice": true, - "max_input_tokens": 131072, - "max_output_tokens": 131072 + "supports_tool_choice": true }, "openrouter/mistralai/mistral-small-3.2-24b-instruct": { "input_cost_per_token": 1e-07, "litellm_provider": "openrouter", + "max_input_tokens": 128000, + "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 3e-07, - "supports_tool_choice": true, - "max_input_tokens": 128000, - "max_output_tokens": 128000 + "supports_tool_choice": true }, "openrouter/mistralai/mixtral-8x22b-instruct": { "input_cost_per_token": 6.5e-07, "litellm_provider": "openrouter", + "max_input_tokens": 65536, + "max_output_tokens": 65536, "max_tokens": 65536, "mode": "chat", "output_cost_per_token": 6.5e-07, - "supports_tool_choice": true, - "max_input_tokens": 65536, - "max_output_tokens": 65536 + "supports_tool_choice": true }, "openrouter/moonshotai/kimi-k2.5": { "cache_read_input_token_cost": 1e-07, @@ -34548,32 +34547,32 @@ "openrouter/openai/gpt-3.5-turbo": { "input_cost_per_token": 1.5e-06, "litellm_provider": "openrouter", + "max_input_tokens": 16385, + "max_output_tokens": 4096, "max_tokens": 4096, "mode": "chat", "output_cost_per_token": 2e-06, - "supports_tool_choice": true, - "max_input_tokens": 16385, - "max_output_tokens": 4096 + "supports_tool_choice": true }, "openrouter/openai/gpt-3.5-turbo-16k": { "input_cost_per_token": 3e-06, "litellm_provider": "openrouter", + "max_input_tokens": 16385, + "max_output_tokens": 4096, "max_tokens": 4096, "mode": "chat", "output_cost_per_token": 4e-06, - "supports_tool_choice": true, - "max_input_tokens": 16385, - "max_output_tokens": 4096 + "supports_tool_choice": true }, "openrouter/openai/gpt-4": { "input_cost_per_token": 3e-05, "litellm_provider": "openrouter", + "max_input_tokens": 8191, + "max_output_tokens": 4096, "max_tokens": 4096, "mode": "chat", "output_cost_per_token": 6e-05, - "supports_tool_choice": true, - "max_input_tokens": 8191, - "max_output_tokens": 4096 + "supports_tool_choice": true }, "openrouter/openai/gpt-4.1": { "cache_read_input_token_cost": 5e-07, @@ -34790,8 +34789,8 @@ "supports_vision": true }, "openrouter/openai/gpt-5.2": { - "input_cost_per_image": 0, "cache_read_input_token_cost": 1.75e-07, + "input_cost_per_image": 0, "input_cost_per_token": 1.75e-06, "litellm_provider": "openrouter", "max_input_tokens": 272000, @@ -34806,8 +34805,8 @@ "supports_vision": true }, "openrouter/openai/gpt-5.2-chat": { - "input_cost_per_image": 0, "cache_read_input_token_cost": 1.75e-07, + "input_cost_per_image": 0, "input_cost_per_token": 1.75e-06, "litellm_provider": "openrouter", "max_input_tokens": 128000, @@ -34940,8 +34939,8 @@ "mode": "chat", "output_cost_per_token": 9.5e-07, "source": "https://openrouter.ai/qwen/qwen3-coder", - "supports_tool_choice": true, - "supports_function_calling": true + "supports_function_calling": true, + "supports_tool_choice": true }, "openrouter/qwen/qwen3-coder-plus": { "input_cost_per_token": 1e-06, @@ -35095,12 +35094,12 @@ "openrouter/undi95/remm-slerp-l2-13b": { "input_cost_per_token": 1.875e-06, "litellm_provider": "openrouter", + "max_input_tokens": 6144, + "max_output_tokens": 4096, "max_tokens": 4096, "mode": "chat", "output_cost_per_token": 1.875e-06, - "supports_tool_choice": true, - "max_input_tokens": 6144, - "max_output_tokens": 4096 + "supports_tool_choice": true }, "openrouter/x-ai/grok-4": { "input_cost_per_token": 3e-06, @@ -35145,89 +35144,89 @@ "supports_tool_choice": true }, "openrouter/xiaomi/mimo-v2-flash": { - "input_cost_per_token": 1e-07, - "output_cost_per_token": 3e-07, "cache_creation_input_token_cost": 0.0, "cache_read_input_token_cost": 1e-08, + "input_cost_per_token": 1e-07, "litellm_provider": "openrouter", "max_input_tokens": 262144, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 3e-07, "supports_function_calling": true, - "supports_tool_choice": true, + "supports_prompt_caching": true, "supports_reasoning": true, - "supports_vision": false, - "supports_prompt_caching": true + "supports_tool_choice": true, + "supports_vision": false }, "openrouter/xiaomi/mimo-v2.5-pro": { - "input_cost_per_token": 1e-06, - "output_cost_per_token": 3e-06, "cache_creation_input_token_cost": 0.0, "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 1e-06, "litellm_provider": "openrouter", "max_input_tokens": 1048576, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 3e-06, "supports_function_calling": true, - "supports_tool_choice": true, + "supports_prompt_caching": true, "supports_reasoning": true, - "supports_vision": false, "supports_response_schema": true, - "supports_prompt_caching": true + "supports_tool_choice": true, + "supports_vision": false }, "openrouter/xiaomi/mimo-v2.5": { - "input_cost_per_token": 4e-07, - "output_cost_per_token": 2e-06, "cache_creation_input_token_cost": 0.0, "cache_read_input_token_cost": 8e-08, + "input_cost_per_token": 4e-07, "litellm_provider": "openrouter", "max_input_tokens": 1048576, "max_output_tokens": 131072, "max_tokens": 131072, "mode": "chat", - "supports_function_calling": true, - "supports_tool_choice": true, - "supports_reasoning": true, - "supports_vision": true, + "output_cost_per_token": 2e-06, "supports_audio_input": true, - "supports_video_input": true, + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, "supports_response_schema": true, - "supports_prompt_caching": true + "supports_tool_choice": true, + "supports_video_input": true, + "supports_vision": true }, "openrouter/z-ai/glm-4.7": { - "input_cost_per_token": 4e-07, - "output_cost_per_token": 1.5e-06, "cache_creation_input_token_cost": 0.0, "cache_read_input_token_cost": 0.0, + "input_cost_per_token": 4e-07, "litellm_provider": "openrouter", "max_input_tokens": 202752, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", + "output_cost_per_token": 1.5e-06, + "supports_assistant_prefill": true, "supports_function_calling": true, - "supports_tool_choice": true, - "supports_reasoning": true, - "supports_vision": true, "supports_prompt_caching": false, - "supports_assistant_prefill": true + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": true }, "openrouter/z-ai/glm-4.7-flash": { - "input_cost_per_token": 7e-08, - "output_cost_per_token": 4e-07, "cache_creation_input_token_cost": 0.0, "cache_read_input_token_cost": 0.0, + "input_cost_per_token": 7e-08, "litellm_provider": "openrouter", "max_input_tokens": 200000, "max_output_tokens": 32000, "max_tokens": 32000, "mode": "chat", + "output_cost_per_token": 4e-07, "supports_function_calling": true, - "supports_tool_choice": true, + "supports_prompt_caching": false, "supports_reasoning": true, - "supports_vision": true, - "supports_prompt_caching": false + "supports_tool_choice": true, + "supports_vision": true }, "openrouter/z-ai/glm-5": { "input_cost_per_token": 8e-07, @@ -35243,15 +35242,15 @@ "supports_tool_choice": true }, "openrouter/z-ai/glm-5.1": { - "input_cost_per_token": 1.05e-06, - "output_cost_per_token": 3.5e-06, - "cache_read_input_token_cost": 5.25e-07, "cache_creation_input_token_cost": 0.0, + "cache_read_input_token_cost": 5.25e-07, + "input_cost_per_token": 1.05e-06, "litellm_provider": "openrouter", "max_input_tokens": 202752, "max_output_tokens": 65535, "max_tokens": 65535, "mode": "chat", + "output_cost_per_token": 3.5e-06, "source": "https://openrouter.ai/z-ai/glm-5.1", "supports_function_calling": true, "supports_prompt_caching": true, @@ -35259,74 +35258,74 @@ "supports_tool_choice": true }, "openrouter/minimax/minimax-m2.1": { - "input_cost_per_token": 2.7e-07, - "output_cost_per_token": 1.2e-06, "cache_creation_input_token_cost": 0.0, "cache_read_input_token_cost": 0.0, + "input_cost_per_token": 2.7e-07, "litellm_provider": "openrouter", "max_input_tokens": 204000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", + "output_cost_per_token": 1.2e-06, + "supports_computer_use": false, "supports_function_calling": true, - "supports_tool_choice": true, - "supports_reasoning": true, - "supports_vision": true, "supports_prompt_caching": false, - "supports_computer_use": false + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": true }, "openrouter/minimax/minimax-m2.5": { - "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.1e-06, "cache_read_input_token_cost": 1.5e-07, + "input_cost_per_token": 3e-07, "litellm_provider": "openrouter", "max_input_tokens": 196608, "max_output_tokens": 65536, "max_tokens": 65536, "mode": "chat", + "output_cost_per_token": 1.1e-06, "source": "https://openrouter.ai/minimax/minimax-m2.5", + "supports_computer_use": false, "supports_function_calling": true, - "supports_tool_choice": true, - "supports_reasoning": true, - "supports_vision": false, "supports_prompt_caching": true, - "supports_computer_use": false + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": false }, "openrouter/openrouter/auto": { "input_cost_per_token": 0, - "output_cost_per_token": 0, "litellm_provider": "openrouter", "max_input_tokens": 2000000, "max_tokens": 2000000, "mode": "chat", + "output_cost_per_token": 0, + "supports_audio_input": true, "supports_function_calling": true, - "supports_tool_choice": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_vision": true, - "supports_audio_input": true, - "supports_video_input": true + "supports_tool_choice": true, + "supports_video_input": true, + "supports_vision": true }, "openrouter/openrouter/free": { "input_cost_per_token": 0, - "output_cost_per_token": 0, "litellm_provider": "openrouter", "max_input_tokens": 200000, "max_tokens": 200000, "mode": "chat", + "output_cost_per_token": 0, "supports_function_calling": true, - "supports_tool_choice": true, "supports_reasoning": true, "supports_response_schema": true, + "supports_tool_choice": true, "supports_vision": true }, "openrouter/openrouter/bodybuilder": { "input_cost_per_token": 0, - "output_cost_per_token": 0, "litellm_provider": "openrouter", "max_input_tokens": 128000, "max_tokens": 128000, - "mode": "chat" + "mode": "chat", + "output_cost_per_token": 0 }, "ovhcloud/DeepSeek-R1-Distill-Llama-70B": { "input_cost_per_token": 6.7e-07, @@ -35890,130 +35889,131 @@ "perplexity/preset/fast-search": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, - "supports_function_calling": true + "supports_function_calling": true, + "supports_web_search": true }, "perplexity/preset/pro-search": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, - "supports_function_calling": true + "supports_function_calling": true, + "supports_web_search": true }, "perplexity/preset/deep-research": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, - "supports_function_calling": true + "supports_function_calling": true, + "supports_web_search": true }, "perplexity/preset/advanced-deep-research": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, - "supports_function_calling": true + "supports_function_calling": true, + "supports_web_search": true }, "perplexity/openai/gpt-5.2": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": true, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/openai/gpt-5.1": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": false, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/openai/gpt-5-mini": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": false, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/anthropic/claude-opus-4-6": { - "supports_adaptive_thinking": true, "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, - "supports_reasoning": false, + "supports_adaptive_thinking": true, "supports_function_calling": true, - "supports_output_config": true + "supports_legacy_thinking": true, + "supports_output_config": true, + "supports_reasoning": false, + "supports_web_search": true }, "perplexity/anthropic/claude-opus-4-7": { - "supports_adaptive_thinking": true, "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, - "supports_reasoning": false, + "supports_adaptive_thinking": true, "supports_function_calling": true, - "supports_output_config": true + "supports_output_config": true, + "supports_reasoning": false, + "supports_web_search": true }, "perplexity/anthropic/claude-opus-4-5": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, - "supports_reasoning": false, "supports_function_calling": true, - "supports_output_config": true + "supports_output_config": true, + "supports_reasoning": false, + "supports_web_search": true }, "perplexity/anthropic/claude-sonnet-4-5": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": false, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/anthropic/claude-haiku-4-5": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": false, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/google/gemini-3-pro-preview": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": false, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/google/gemini-3-flash-preview": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": false, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/google/gemini-2.5-pro": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": false, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/google/gemini-2.5-flash": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, - "supports_reasoning": false, "supports_function_calling": true, - "supports_image_size": false + "supports_image_size": false, + "supports_reasoning": false, + "supports_web_search": true }, "perplexity/xai/grok-4-1-fast-non-reasoning": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": false, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/perplexity/sonar": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": false, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/perplexity/deepseek-v4-flash-0731": { "cache_read_input_token_cost": 2.8e-08, @@ -36022,9 +36022,9 @@ "mode": "responses", "output_cost_per_token": 2.6e-07, "source": "https://docs.perplexity.ai/docs/agent-api/models", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": true, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/perplexity/glm-5.2": { "cache_read_input_token_cost": 1.4e-07, @@ -36033,9 +36033,9 @@ "mode": "responses", "output_cost_per_token": 4.4e-06, "source": "https://docs.perplexity.ai/docs/agent-api/models", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": true, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/perplexity/kimi-k3": { "cache_read_input_token_cost": 3e-07, @@ -36044,9 +36044,9 @@ "mode": "responses", "output_cost_per_token": 1.5e-05, "source": "https://docs.perplexity.ai/docs/agent-api/models", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": true, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/perplexity/kimi-k2.7-code": { "cache_read_input_token_cost": 1.9e-07, @@ -36055,9 +36055,9 @@ "mode": "responses", "output_cost_per_token": 4e-06, "source": "https://docs.perplexity.ai/docs/agent-api/models", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": false, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/pplx-embed-v1-0.6b": { "input_cost_per_token": 4e-09, @@ -36101,8 +36101,8 @@ "output_cost_per_token": 0.0, "source": "https://platform.publicai.co/docs", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_reasoning": true + "supports_reasoning": true, + "supports_tool_choice": true }, "publicai/allenai/Olmo-3-32B-Think": { "input_cost_per_token": 0.0, @@ -36114,8 +36114,8 @@ "output_cost_per_token": 0.0, "source": "https://platform.publicai.co/docs", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_reasoning": true + "supports_reasoning": true, + "supports_tool_choice": true }, "qwen.qwen3-coder-480b-a35b-v1:0": { "input_cost_per_token": 2.2e-07, @@ -36126,24 +36126,24 @@ "mode": "chat", "output_cost_per_token": 1.8e-06, "supports_function_calling": true, + "supports_native_structured_output": true, "supports_reasoning": true, - "supports_tool_choice": true, - "supports_native_structured_output": true + "supports_tool_choice": true }, "qwen.qwen3-235b-a22b-2507-v1:0": { "input_cost_per_token": 2.2e-07, + "input_cost_per_token_batches": 1.1e-07, "litellm_provider": "bedrock_converse", "max_input_tokens": 262144, "max_output_tokens": 131072, "max_tokens": 131072, "mode": "chat", "output_cost_per_token": 8.8e-07, + "output_cost_per_token_batches": 4.4e-07, "supports_function_calling": true, - "supports_reasoning": true, - "supports_tool_choice": true, "supports_native_structured_output": true, - "input_cost_per_token_batches": 1.1e-07, - "output_cost_per_token_batches": 4.4e-07 + "supports_reasoning": true, + "supports_tool_choice": true }, "qwen.qwen3-coder-30b-a3b-v1:0": { "input_cost_per_token": 1.5e-07, @@ -36154,9 +36154,9 @@ "mode": "chat", "output_cost_per_token": 6e-07, "supports_function_calling": true, + "supports_native_structured_output": true, "supports_reasoning": true, - "supports_tool_choice": true, - "supports_native_structured_output": true + "supports_tool_choice": true }, "qwen.qwen3-32b-v1:0": { "input_cost_per_token": 1.5e-07, @@ -36167,9 +36167,9 @@ "mode": "chat", "output_cost_per_token": 6e-07, "supports_function_calling": true, + "supports_native_structured_output": true, "supports_reasoning": true, - "supports_tool_choice": true, - "supports_native_structured_output": true + "supports_tool_choice": true }, "qwen.qwen3-next-80b-a3b": { "input_cost_per_token": 1.5e-07, @@ -36180,8 +36180,8 @@ "mode": "chat", "output_cost_per_token": 1.2e-06, "supports_function_calling": true, - "supports_system_messages": true, - "supports_native_structured_output": true + "supports_native_structured_output": true, + "supports_system_messages": true }, "qwen.qwen3-vl-235b-a22b": { "input_cost_per_token": 5.3e-07, @@ -36192,9 +36192,9 @@ "mode": "chat", "output_cost_per_token": 2.66e-06, "supports_function_calling": true, + "supports_native_structured_output": true, "supports_system_messages": true, - "supports_vision": true, - "supports_native_structured_output": true + "supports_vision": true }, "qwen.qwen3-coder-next": { "input_cost_per_token": 5e-07, @@ -36204,10 +36204,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "reducto/parse-legacy": { "litellm_provider": "reducto", @@ -36377,300 +36377,301 @@ }, "replicate/openai/gpt-5": { "input_cost_per_token": 1.25e-06, - "output_cost_per_token": 1e-05, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 1e-05, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "replicate/openai/gpt-oss-20b": { "input_cost_per_token": 9e-08, - "output_cost_per_token": 3.6e-07, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 3.6e-07, "supports_function_calling": true, "supports_system_messages": true }, "replicate/anthropic/claude-4.5-haiku": { "input_cost_per_token": 1e-06, - "output_cost_per_token": 5e-06, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 5e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_prompt_caching": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true, - "supports_prompt_caching": true + "supports_vision": true }, "replicate/ibm-granite/granite-3.3-8b-instruct": { "input_cost_per_token": 3e-08, - "output_cost_per_token": 2.5e-07, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 2.5e-07, "supports_function_calling": true, "supports_system_messages": true }, "replicate/openai/gpt-4o": { "input_cost_per_token": 2.5e-06, - "output_cost_per_token": 1e-05, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 1e-05, + "supports_audio_input": true, + "supports_audio_output": true, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true, - "supports_audio_input": true, - "supports_audio_output": true + "supports_vision": true }, "replicate/openai/o4-mini": { "input_cost_per_token": 1e-06, - "output_cost_per_token": 4e-06, - "output_cost_per_reasoning_token": 4e-06, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_reasoning_token": 4e-06, + "output_cost_per_token": 4e-06, "supports_reasoning": true, "supports_system_messages": true }, "replicate/openai/o1-mini": { "input_cost_per_token": 1.1e-06, - "output_cost_per_token": 4.4e-06, - "output_cost_per_reasoning_token": 4.4e-06, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_reasoning_token": 4.4e-06, + "output_cost_per_token": 4.4e-06, "supports_reasoning": true, "supports_system_messages": true }, "replicate/openai/o1": { "input_cost_per_token": 1.5e-05, - "output_cost_per_token": 6e-05, - "output_cost_per_reasoning_token": 6e-05, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_reasoning_token": 6e-05, + "output_cost_per_token": 6e-05, "supports_reasoning": true, "supports_system_messages": true }, "replicate/openai/gpt-4o-mini": { "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 6e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "replicate/qwen/qwen3-235b-a22b-instruct-2507": { "input_cost_per_token": 2.64e-07, - "output_cost_per_token": 1.06e-06, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 1.06e-06, "supports_function_calling": true, "supports_system_messages": true }, "replicate/anthropic/claude-4-sonnet": { "input_cost_per_token": 3e-06, - "output_cost_per_token": 1.5e-05, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 1.5e-05, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_prompt_caching": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true, - "supports_prompt_caching": true + "supports_vision": true }, "replicate/deepseek-ai/deepseek-v3": { "input_cost_per_token": 1.45e-06, - "output_cost_per_token": 1.45e-06, "litellm_provider": "replicate", - "mode": "chat", "max_input_tokens": 65536, "max_output_tokens": 8192, "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 1.45e-06, "supports_function_calling": true, "supports_system_messages": true }, "replicate/anthropic/claude-3.7-sonnet": { "input_cost_per_token": 3e-06, - "output_cost_per_token": 1.5e-05, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 1.5e-05, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_prompt_caching": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true, - "supports_prompt_caching": true + "supports_vision": true }, "replicate/anthropic/claude-3.5-haiku": { "input_cost_per_token": 1e-06, - "output_cost_per_token": 5e-06, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 5e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_prompt_caching": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true, - "supports_prompt_caching": true + "supports_vision": true }, "replicate/anthropic/claude-3.5-sonnet": { "input_cost_per_token": 3.75e-06, - "output_cost_per_token": 1.875e-05, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 1.875e-05, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_prompt_caching": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true, - "supports_prompt_caching": true + "supports_vision": true }, "replicate/google/gemini-3-pro": { "input_cost_per_token": 2e-06, - "output_cost_per_token": 1.2e-05, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 1.2e-05, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "replicate/anthropic/claude-4.5-sonnet": { "input_cost_per_token": 3e-06, - "output_cost_per_token": 1.5e-05, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 1.5e-05, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_prompt_caching": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true, - "supports_prompt_caching": true + "supports_vision": true }, "replicate/openai/gpt-4.1": { "input_cost_per_token": 2e-06, - "output_cost_per_token": 8e-06, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 8e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "replicate/openai/gpt-4.1-nano": { "input_cost_per_token": 1e-07, - "output_cost_per_token": 4e-07, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 4e-07, "supports_function_calling": true, "supports_system_messages": true }, "replicate/openai/gpt-4.1-mini": { "input_cost_per_token": 4e-07, - "output_cost_per_token": 1.6e-06, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 1.6e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "replicate/openai/gpt-5-nano": { "input_cost_per_token": 5e-08, - "output_cost_per_token": 4e-07, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 4e-07, "supports_function_calling": true, "supports_system_messages": true }, "replicate/openai/gpt-5-mini": { "input_cost_per_token": 2.5e-07, - "output_cost_per_token": 2e-06, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 2e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "replicate/google/gemini-2.5-flash": { "input_cost_per_token": 2.5e-06, - "output_cost_per_token": 2.5e-06, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 2.5e-06, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true, - "supports_image_size": false + "supports_vision": true }, "replicate/openai/gpt-oss-120b": { "input_cost_per_token": 1.8e-07, - "output_cost_per_token": 7.2e-07, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 7.2e-07, "supports_function_calling": true, "supports_system_messages": true }, "replicate/deepseek-ai/deepseek-v3.1": { "input_cost_per_token": 6.72e-07, - "output_cost_per_token": 2.016e-06, "litellm_provider": "replicate", - "mode": "chat", "max_input_tokens": 163840, "max_output_tokens": 163840, "max_tokens": 163840, + "mode": "chat", + "output_cost_per_token": 2.016e-06, "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true }, "replicate/xai/grok-4": { "input_cost_per_token": 7.2e-06, - "output_cost_per_token": 3.6e-05, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 3.6e-05, "supports_function_calling": true, "supports_system_messages": true }, "replicate/deepseek-ai/deepseek-r1": { "input_cost_per_token": 3.75e-06, - "output_cost_per_token": 1e-05, - "output_cost_per_reasoning_token": 1e-05, "litellm_provider": "replicate", - "mode": "chat", "max_input_tokens": 65536, "max_output_tokens": 8192, "max_tokens": 8192, + "mode": "chat", + "output_cost_per_reasoning_token": 1e-05, + "output_cost_per_token": 1e-05, "supports_reasoning": true, "supports_system_messages": true }, "rerank-english-v2.0": { + "deprecation_date": "2025-04-30", "input_cost_per_query": 0.002, "input_cost_per_token": 0.0, "litellm_provider": "cohere", @@ -36678,8 +36679,7 @@ "max_output_tokens": 4096, "max_tokens": 4096, "mode": "rerank", - "output_cost_per_token": 0.0, - "deprecation_date": "2025-04-30" + "output_cost_per_token": 0.0 }, "rerank-english-v3.0": { "input_cost_per_query": 0.002, @@ -36692,6 +36692,7 @@ "output_cost_per_token": 0.0 }, "rerank-multilingual-v2.0": { + "deprecation_date": "2025-04-30", "input_cost_per_query": 0.002, "input_cost_per_token": 0.0, "litellm_provider": "cohere", @@ -36699,8 +36700,7 @@ "max_output_tokens": 4096, "max_tokens": 4096, "mode": "rerank", - "output_cost_per_token": 0.0, - "deprecation_date": "2025-04-30" + "output_cost_per_token": 0.0 }, "rerank-multilingual-v3.0": { "input_cost_per_query": 0.002, @@ -36991,57 +36991,57 @@ "supports_tool_choice": true }, "sambanova/DeepSeek-V3.1": { - "max_tokens": 131072, + "input_cost_per_token": 3e-06, + "litellm_provider": "sambanova", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 3e-06, - "output_cost_per_token": 4.5e-06, - "litellm_provider": "sambanova", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 4.5e-06, + "source": "https://cloud.sambanova.ai/plans/pricing", "supports_function_calling": true, - "supports_tool_choice": true, "supports_reasoning": true, - "source": "https://cloud.sambanova.ai/plans/pricing" + "supports_tool_choice": true }, "sambanova/gpt-oss-120b": { - "max_tokens": 131072, + "input_cost_per_token": 2.2e-07, + "litellm_provider": "sambanova", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2.2e-07, - "output_cost_per_token": 5.9e-07, - "litellm_provider": "sambanova", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 5.9e-07, + "source": "https://cloud.sambanova.ai/plans/pricing", "supports_function_calling": true, - "supports_tool_choice": true, "supports_reasoning": true, - "source": "https://cloud.sambanova.ai/plans/pricing" + "supports_tool_choice": true }, "sambanova/DeepSeek-V3.2": { - "max_tokens": 32768, + "input_cost_per_token": 3e-06, + "litellm_provider": "sambanova", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 3e-06, - "output_cost_per_token": 4.5e-06, - "litellm_provider": "sambanova", + "max_tokens": 32768, "mode": "chat", - "supports_function_calling": true, - "supports_tool_choice": true, + "output_cost_per_token": 4.5e-06, "source": "https://cloud.sambanova.ai/plans/pricing", "supported_endpoints": [ "/v1/chat/completions" ], + "supports_function_calling": true, + "supports_tool_choice": true, "supports_vision": false }, "sambanova/gemma-4-31B-it": { - "max_tokens": 131072, + "input_cost_per_token": 3.8e-07, + "litellm_provider": "sambanova", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 3.8e-07, - "output_cost_per_token": 1.15e-06, - "litellm_provider": "sambanova", + "max_tokens": 131072, "mode": "chat", - "supports_vision": true, - "source": "https://cloud.sambanova.ai/plans/pricing" + "output_cost_per_token": 1.15e-06, + "source": "https://cloud.sambanova.ai/plans/pricing", + "supports_vision": true }, "scx-ai/GLM-5.2": { "cache_read_input_token_cost": 2.2e-07, @@ -37078,28 +37078,28 @@ "supports_vision": true }, "snowflake/claude-3-5-sonnet": { + "cache_read_input_token_cost": 3e-07, + "input_cost_per_token": 3e-06, "litellm_provider": "snowflake", "max_input_tokens": 200000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 3e-06, "output_cost_per_token": 1.5e-05, - "cache_read_input_token_cost": 3e-07, "supports_computer_use": true, "supports_function_calling": true, - "supports_vision": true, "supports_prompt_caching": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_vision": true }, "snowflake/deepseek-r1": { + "input_cost_per_token": 1.35e-06, "litellm_provider": "snowflake", "max_input_tokens": 128000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 1.35e-06, "output_cost_per_token": 5.4e-06, "supports_reasoning": true, "supports_system_messages": true @@ -37154,34 +37154,34 @@ "mode": "chat" }, "snowflake/llama3.1-405b": { + "input_cost_per_token": 1.2e-06, "litellm_provider": "snowflake", "max_input_tokens": 128000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 1.2e-06, "output_cost_per_token": 1.2e-06, "supports_function_calling": true, "supports_system_messages": true }, "snowflake/llama3.1-70b": { + "input_cost_per_token": 7.2e-07, "litellm_provider": "snowflake", "max_input_tokens": 128000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 7.2e-07, "output_cost_per_token": 7.2e-07, "supports_function_calling": true, "supports_system_messages": true }, "snowflake/llama3.1-8b": { + "input_cost_per_token": 2.4e-07, "litellm_provider": "snowflake", "max_input_tokens": 128000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 2.4e-07, "output_cost_per_token": 2.4e-07, "supports_system_messages": true }, @@ -37200,13 +37200,13 @@ "mode": "chat" }, "snowflake/llama3.3-70b": { - "max_tokens": 16384, + "input_cost_per_token": 7.2e-07, + "litellm_provider": "snowflake", "max_input_tokens": 128000, "max_output_tokens": 16384, - "input_cost_per_token": 7.2e-07, - "output_cost_per_token": 7.2e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 7.2e-07, "supports_function_calling": true, "supports_system_messages": true }, @@ -37225,16 +37225,16 @@ "mode": "chat" }, "snowflake/mistral-large2": { + "input_cost_per_token": 2e-06, "litellm_provider": "snowflake", "max_input_tokens": 128000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 2e-06, "output_cost_per_token": 6e-06, "supports_function_calling": true, - "supports_system_messages": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_system_messages": true }, "snowflake/mixtral-8x7b": { "litellm_provider": "snowflake", @@ -37272,13 +37272,13 @@ "mode": "chat" }, "snowflake/snowflake-llama-3.3-70b": { - "max_tokens": 16384, + "input_cost_per_token": 7.2e-07, + "litellm_provider": "snowflake", "max_input_tokens": 128000, "max_output_tokens": 16384, - "input_cost_per_token": 7.2e-07, - "output_cost_per_token": 7.2e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 7.2e-07, "supports_function_calling": true, "supports_system_messages": true }, @@ -37969,6 +37969,8 @@ "together_ai/deepseek-ai/DeepSeek-V3.1": { "input_cost_per_token": 6e-07, "litellm_provider": "together_ai", + "max_input_tokens": 128000, + "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", "output_cost_per_token": 1.7e-06, @@ -37976,9 +37978,7 @@ "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_reasoning": true, - "supports_tool_choice": true, - "max_input_tokens": 128000, - "max_output_tokens": 16384 + "supports_tool_choice": true }, "together_ai/meta-llama/Llama-3.2-3B-Instruct-Turbo": { "litellm_provider": "together_ai", @@ -38178,9 +38178,9 @@ "output_cost_per_token": 2.8e-06, "source": "https://www.together.ai/models/kimi-k2-5", "supports_function_calling": true, + "supports_reasoning": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_reasoning": true + "supports_vision": true }, "together_ai/moonshotai/Kimi-K2-Instruct-0905": { "input_cost_per_token": 1e-06, @@ -38249,37 +38249,37 @@ "input_cost_per_character": 4e-06, "litellm_provider": "aws_polly", "mode": "audio_speech", + "source": "https://aws.amazon.com/polly/pricing/", "supported_endpoints": [ "/v1/audio/speech" - ], - "source": "https://aws.amazon.com/polly/pricing/" + ] }, "aws_polly/neural": { "input_cost_per_character": 1.6e-05, "litellm_provider": "aws_polly", "mode": "audio_speech", + "source": "https://aws.amazon.com/polly/pricing/", "supported_endpoints": [ "/v1/audio/speech" - ], - "source": "https://aws.amazon.com/polly/pricing/" + ] }, "aws_polly/long-form": { "input_cost_per_character": 0.0001, "litellm_provider": "aws_polly", "mode": "audio_speech", + "source": "https://aws.amazon.com/polly/pricing/", "supported_endpoints": [ "/v1/audio/speech" - ], - "source": "https://aws.amazon.com/polly/pricing/" + ] }, "aws_polly/generative": { "input_cost_per_character": 3e-05, "litellm_provider": "aws_polly", "mode": "audio_speech", + "source": "https://aws.amazon.com/polly/pricing/", "supported_endpoints": [ "/v1/audio/speech" - ], - "source": "https://aws.amazon.com/polly/pricing/" + ] }, "us.amazon.nova-lite-v1:0": { "input_cost_per_token": 6e-08, @@ -38346,43 +38346,45 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 4e-06, + "prompt_cache_min_tokens": 2048, "supports_assistant_prefill": true, "supports_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_response_schema": true, - "supports_tool_choice": true, - "prompt_cache_min_tokens": 2048 + "supports_tool_choice": true }, "us.anthropic.claude-haiku-4-5-20251001-v1:0": { "cache_creation_input_token_cost": 1.375e-06, "cache_creation_input_token_cost_above_1hr": 2.2e-06, "cache_read_input_token_cost": 1.1e-07, "input_cost_per_token": 1.1e-06, + "input_cost_per_token_batches": 5.5e-07, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5.5e-06, + "output_cost_per_token_batches": 2.75e-06, + "prompt_cache_min_tokens": 4096, "source": "https://aws.amazon.com/about-aws/whats-new/2025/10/claude-4-5-haiku-anthropic-amazon-bedrock", "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096, - "input_cost_per_token_batches": 5.5e-07, - "output_cost_per_token_batches": 2.75e-06 + "supports_tool_search": true, + "supports_vision": true }, "us.anthropic.claude-3-5-sonnet-20240620-v1:0": { + "cache_creation_input_token_cost": 3.75e-06, + "cache_read_input_token_cost": 3e-07, "input_cost_per_token": 3e-06, "litellm_provider": "bedrock", "max_input_tokens": 200000, @@ -38394,9 +38396,7 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-07, - "cache_creation_input_token_cost": 3.75e-06 + "supports_vision": true }, "us.anthropic.claude-3-5-sonnet-20241022-v2:0": { "cache_creation_input_token_cost": 3.75e-06, @@ -38438,6 +38438,8 @@ "supports_vision": true }, "us.anthropic.claude-3-haiku-20240307-v1:0": { + "cache_creation_input_token_cost": 3.125e-07, + "cache_read_input_token_cost": 2.5e-08, "deprecation_date": "2026-09-10", "input_cost_per_token": 2.5e-07, "litellm_provider": "bedrock", @@ -38450,11 +38452,11 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 2.5e-08, - "cache_creation_input_token_cost": 3.125e-07 + "supports_vision": true }, "us.anthropic.claude-3-opus-20240229-v1:0": { + "cache_creation_input_token_cost": 1.875e-05, + "cache_read_input_token_cost": 1.5e-06, "input_cost_per_token": 1.5e-05, "litellm_provider": "bedrock", "max_input_tokens": 200000, @@ -38465,11 +38467,11 @@ "supports_function_calling": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 1.5e-06, - "cache_creation_input_token_cost": 1.875e-05 + "supports_vision": true }, "us.anthropic.claude-3-sonnet-20240229-v1:0": { + "cache_creation_input_token_cost": 3.75e-06, + "cache_read_input_token_cost": 3e-07, "deprecation_date": "2026-07-30", "input_cost_per_token": 3e-06, "litellm_provider": "bedrock", @@ -38482,9 +38484,7 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-07, - "cache_creation_input_token_cost": 3.75e-06 + "supports_vision": true }, "us.anthropic.claude-opus-4-1-20250805-v1:0": { "cache_creation_input_token_cost": 1.875e-05, @@ -38497,6 +38497,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -38510,26 +38511,27 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "us.anthropic.claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 4.125e-06, "cache_creation_input_token_cost_above_1hr": 6.6e-06, + "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.32e-05, + "cache_creation_input_token_cost_above_200k_tokens": 8.25e-06, "cache_read_input_token_cost": 3.3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6.6e-07, "input_cost_per_token": 3.3e-06, "input_cost_per_token_above_200k_tokens": 6.6e-06, - "output_cost_per_token_above_200k_tokens": 2.475e-05, - "cache_creation_input_token_cost_above_200k_tokens": 8.25e-06, - "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.32e-05, - "cache_read_input_token_cost_above_200k_tokens": 6.6e-07, + "input_cost_per_token_batches": 1.65e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.65e-05, + "output_cost_per_token_above_200k_tokens": 2.475e-05, + "output_cost_per_token_batches": 8.25e-06, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -38538,74 +38540,72 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024, - "input_cost_per_token_batches": 1.65e-06, - "output_cost_per_token_batches": 8.25e-06 + "supports_tool_search": true, + "supports_vision": true }, "us-gov.anthropic.claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 4.5e-06, "cache_creation_input_token_cost_above_1hr": 7.2e-06, + "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.44e-05, + "cache_creation_input_token_cost_above_200k_tokens": 9e-06, "cache_read_input_token_cost": 3.6e-07, + "cache_read_input_token_cost_above_200k_tokens": 7.2e-07, "input_cost_per_token": 3.6e-06, "input_cost_per_token_above_200k_tokens": 7.2e-06, - "output_cost_per_token_above_200k_tokens": 2.7e-05, - "cache_creation_input_token_cost_above_200k_tokens": 9e-06, - "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.44e-05, - "cache_read_input_token_cost_above_200k_tokens": 7.2e-07, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.8e-05, + "output_cost_per_token_above_200k_tokens": 2.7e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_tool_search": true, + "supports_vision": true }, "au.anthropic.claude-haiku-4-5-20251001-v1:0": { "cache_creation_input_token_cost": 1.375e-06, "cache_creation_input_token_cost_above_1hr": 2.2e-06, "cache_read_input_token_cost": 1.1e-07, "input_cost_per_token": 1.1e-06, + "input_cost_per_token_batches": 5.5e-07, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5.5e-06, + "output_cost_per_token_batches": 2.75e-06, + "prompt_cache_min_tokens": 4096, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096, - "input_cost_per_token_batches": 5.5e-07, - "output_cost_per_token_batches": 2.75e-06 + "supports_tool_search": true, + "supports_vision": true }, "us.anthropic.claude-opus-4-20250514-v1:0": { "cache_creation_input_token_cost": 1.875e-05, @@ -38617,6 +38617,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -38630,21 +38631,21 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "us.anthropic.claude-opus-4-5-20251101-v1:0": { + "bedrock_output_config_effort_ceiling": "high", "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -38653,30 +38654,30 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "high", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "global.anthropic.claude-opus-4-5-20251101-v1:0": { + "bedrock_output_config_effort_ceiling": "high", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -38685,29 +38686,29 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "high", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "eu.anthropic.claude-opus-4-5-20251101-v1:0": { + "bedrock_output_config_effort_ceiling": "high", "cache_creation_input_token_cost": 6.25e-06, "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -38716,33 +38717,34 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "high", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "us.anthropic.claude-sonnet-4-20250514-v1:0": { + "bedrock_converse_supports_strict_tools": false, "cache_creation_input_token_cost": 3.75e-06, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, "deprecation_date": "2026-10-14", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -38756,9 +38758,7 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "bedrock_converse_supports_strict_tools": false, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "us.deepseek.r1-v1:0": { "input_cost_per_token": 1.35e-06, @@ -38781,8 +38781,8 @@ "mode": "chat", "output_cost_per_token": 1.85e-06, "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, + "supports_reasoning": true, "supports_tool_choice": true }, "eu.deepseek.v3.2": { @@ -38794,8 +38794,8 @@ "mode": "chat", "output_cost_per_token": 2.22e-06, "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, + "supports_reasoning": true, "supports_tool_choice": true }, "us.meta.llama3-1-405b-instruct-v1:0": { @@ -39040,9 +39040,9 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 2.4e-07, - "supports_vision": true, "supports_function_calling": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_vision": true }, "vercel_ai_gateway/amazon/nova-micro": { "input_cost_per_token": 3.5e-08, @@ -39063,9 +39063,9 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 3.2e-06, - "supports_vision": true, "supports_function_calling": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_vision": true }, "vercel_ai_gateway/amazon/titan-embed-text-v2": { "input_cost_per_token": 2e-08, @@ -39086,10 +39086,10 @@ "max_tokens": 4096, "mode": "chat", "output_cost_per_token": 1.25e-06, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/anthropic/claude-3-opus": { "cache_creation_input_token_cost": 1.875e-05, @@ -39101,10 +39101,10 @@ "max_tokens": 4096, "mode": "chat", "output_cost_per_token": 7.5e-05, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/anthropic/claude-3.5-haiku": { "cache_creation_input_token_cost": 1e-06, @@ -39116,10 +39116,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 4e-06, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/anthropic/claude-3.5-sonnet": { "cache_creation_input_token_cost": 3.75e-06, @@ -39131,10 +39131,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.5e-05, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/anthropic/claude-3.7-sonnet": { "cache_creation_input_token_cost": 3.75e-06, @@ -39146,10 +39146,10 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/anthropic/claude-4-opus": { "cache_creation_input_token_cost": 1.875e-05, @@ -39161,10 +39161,10 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/anthropic/claude-4-sonnet": { "cache_creation_input_token_cost": 3.75e-06, @@ -39244,6 +39244,7 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, + "prompt_cache_min_tokens": 4096, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, @@ -39251,8 +39252,7 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "vercel_ai_gateway/anthropic/claude-opus-4": { "cache_creation_input_token_cost": 1.875e-05, @@ -39283,6 +39283,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, @@ -39290,8 +39291,7 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "vercel_ai_gateway/anthropic/claude-opus-4.5": { "cache_creation_input_token_cost": 6.25e-06, @@ -39303,19 +39303,18 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_output_config": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "vercel_ai_gateway/anthropic/claude-opus-4.6": { - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, @@ -39325,16 +39324,18 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, + "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_output_config": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "vercel_ai_gateway/anthropic/claude-sonnet-4": { "cache_creation_input_token_cost": 3.75e-06, @@ -39346,6 +39347,7 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, @@ -39353,8 +39355,7 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "vercel_ai_gateway/anthropic/claude-sonnet-4.5": { "cache_creation_input_token_cost": 3.75e-06, @@ -39366,14 +39367,14 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "vercel_ai_gateway/cohere/command-a": { "input_cost_per_token": 2.5e-06, @@ -39384,8 +39385,8 @@ "mode": "chat", "output_cost_per_token": 1e-05, "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "vercel_ai_gateway/cohere/command-r": { "input_cost_per_token": 1.5e-07, @@ -39437,8 +39438,8 @@ "mode": "chat", "output_cost_per_token": 9.9e-07, "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "vercel_ai_gateway/deepseek/deepseek-v3": { "input_cost_per_token": 9e-07, @@ -39459,10 +39460,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 6e-07, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/google/gemini-2.0-flash-lite": { "deprecation_date": "2026-06-01", @@ -39473,10 +39474,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 3e-07, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/google/gemini-2.5-flash": { "input_cost_per_token": 3e-07, @@ -39486,11 +39487,11 @@ "max_tokens": 65536, "mode": "chat", "output_cost_per_token": 2.5e-06, - "supports_vision": true, "supports_function_calling": true, - "supports_tool_choice": true, + "supports_image_size": false, "supports_response_schema": true, - "supports_image_size": false + "supports_tool_choice": true, + "supports_vision": true }, "vercel_ai_gateway/google/gemini-2.5-pro": { "input_cost_per_token": 2.5e-06, @@ -39500,10 +39501,10 @@ "max_tokens": 65536, "mode": "chat", "output_cost_per_token": 1e-05, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/google/gemini-embedding-001": { "input_cost_per_token": 1.5e-07, @@ -39522,9 +39523,9 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 2e-07, - "supports_vision": true, "supports_function_calling": true, - "supports_tool_choice": true + "supports_tool_choice": true, + "supports_vision": true }, "vercel_ai_gateway/google/text-embedding-005": { "input_cost_per_token": 2.5e-08, @@ -39602,9 +39603,9 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.6e-07, - "supports_vision": true, "supports_function_calling": true, - "supports_tool_choice": true + "supports_tool_choice": true, + "supports_vision": true }, "vercel_ai_gateway/meta/llama-3.2-1b": { "input_cost_per_token": 1e-07, @@ -39634,9 +39635,9 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 7.2e-07, - "supports_vision": true, "supports_function_calling": true, - "supports_tool_choice": true + "supports_tool_choice": true, + "supports_vision": true }, "vercel_ai_gateway/meta/llama-3.3-70b": { "input_cost_per_token": 7.2e-07, @@ -39667,9 +39668,9 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 3e-07, - "supports_vision": true, "supports_function_calling": true, - "supports_tool_choice": true + "supports_tool_choice": true, + "supports_vision": true }, "vercel_ai_gateway/mistral/codestral": { "input_cost_per_token": 3e-07, @@ -39700,8 +39701,8 @@ "mode": "chat", "output_cost_per_token": 2.8e-07, "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "vercel_ai_gateway/mistral/magistral-medium": { "input_cost_per_token": 2e-06, @@ -39712,8 +39713,8 @@ "mode": "chat", "output_cost_per_token": 5e-06, "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "vercel_ai_gateway/mistral/magistral-small": { "input_cost_per_token": 5e-07, @@ -39744,9 +39745,9 @@ "max_tokens": 4000, "mode": "chat", "output_cost_per_token": 1e-07, - "supports_vision": true, "supports_function_calling": true, - "supports_tool_choice": true + "supports_tool_choice": true, + "supports_vision": true }, "vercel_ai_gateway/mistral/mistral-embed": { "input_cost_per_token": 1e-07, @@ -39786,8 +39787,8 @@ "mode": "chat", "output_cost_per_token": 3e-07, "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "vercel_ai_gateway/mistral/mixtral-8x22b-instruct": { "input_cost_per_token": 1.2e-06, @@ -39807,10 +39808,10 @@ "max_tokens": 4000, "mode": "chat", "output_cost_per_token": 1.5e-07, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/mistral/pixtral-large": { "input_cost_per_token": 2e-06, @@ -39820,10 +39821,10 @@ "max_tokens": 4000, "mode": "chat", "output_cost_per_token": 6e-06, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/moonshotai/kimi-k2": { "input_cost_per_token": 5.5e-07, @@ -39882,9 +39883,9 @@ "max_tokens": 4096, "mode": "chat", "output_cost_per_token": 3e-05, - "supports_vision": true, "supports_function_calling": true, - "supports_tool_choice": true + "supports_tool_choice": true, + "supports_vision": true }, "vercel_ai_gateway/openai/gpt-4.1": { "cache_creation_input_token_cost": 0.0, @@ -39896,10 +39897,10 @@ "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 8e-06, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/openai/gpt-4.1-mini": { "cache_creation_input_token_cost": 0.0, @@ -39911,10 +39912,10 @@ "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 1.6e-06, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/openai/gpt-4.1-nano": { "cache_creation_input_token_cost": 0.0, @@ -39926,10 +39927,10 @@ "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 4e-07, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/openai/gpt-4o": { "cache_creation_input_token_cost": 0.0, @@ -39941,10 +39942,10 @@ "max_tokens": 16384, "mode": "chat", "output_cost_per_token": 1e-05, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/openai/gpt-4o-mini": { "cache_creation_input_token_cost": 0.0, @@ -39956,10 +39957,10 @@ "max_tokens": 16384, "mode": "chat", "output_cost_per_token": 6e-07, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/openai/o1": { "cache_creation_input_token_cost": 0.0, @@ -39971,10 +39972,10 @@ "max_tokens": 100000, "mode": "chat", "output_cost_per_token": 6e-05, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/openai/o3": { "cache_creation_input_token_cost": 0.0, @@ -39986,10 +39987,10 @@ "max_tokens": 100000, "mode": "chat", "output_cost_per_token": 8e-06, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/openai/o3-mini": { "cache_creation_input_token_cost": 0.0, @@ -40002,8 +40003,8 @@ "mode": "chat", "output_cost_per_token": 4.4e-06, "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "vercel_ai_gateway/openai/o4-mini": { "cache_creation_input_token_cost": 0.0, @@ -40015,10 +40016,10 @@ "max_tokens": 100000, "mode": "chat", "output_cost_per_token": 4.4e-06, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/openai/text-embedding-3-large": { "input_cost_per_token": 1.3e-07, @@ -40091,9 +40092,9 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 1.5e-05, - "supports_vision": true, "supports_function_calling": true, - "supports_tool_choice": true + "supports_tool_choice": true, + "supports_vision": true }, "vercel_ai_gateway/vercel/v0-1.5-md": { "input_cost_per_token": 3e-06, @@ -40103,9 +40104,9 @@ "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 1.5e-05, - "supports_vision": true, "supports_function_calling": true, - "supports_tool_choice": true + "supports_tool_choice": true, + "supports_vision": true }, "vercel_ai_gateway/xai/grok-2": { "input_cost_per_token": 2e-06, @@ -40126,9 +40127,9 @@ "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 1e-05, - "supports_vision": true, "supports_function_calling": true, - "supports_tool_choice": true + "supports_tool_choice": true, + "supports_vision": true }, "vercel_ai_gateway/xai/grok-3": { "input_cost_per_token": 3e-06, @@ -40207,9 +40208,9 @@ "supports_tool_choice": true }, "vercel_ai_gateway/zai/glm-4.6": { - "litellm_provider": "vercel_ai_gateway", "cache_read_input_token_cost": 1.1e-07, "input_cost_per_token": 4.5e-07, + "litellm_provider": "vercel_ai_gateway", "max_input_tokens": 200000, "max_output_tokens": 200000, "max_tokens": 200000, @@ -40269,10 +40270,10 @@ "supports_tool_choice": true }, "vertex_ai/claude-haiku-4-5": { - "deprecation_date": "2026-10-15", "cache_creation_input_token_cost": 1.25e-06, "cache_creation_input_token_cost_above_1hr": 2e-06, "cache_read_input_token_cost": 1e-07, + "deprecation_date": "2026-10-15", "input_cost_per_token": 1e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 200000, @@ -40280,24 +40281,24 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, + "prompt_cache_min_tokens": 4096, "regional_endpoint_uplift_multiplier": 1.1, "source": "https://cloud.google.com/vertex-ai/generative-ai/docs/partner-models/claude/haiku-4-5", "supports_assistant_prefill": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_native_streaming": true, - "supports_vision": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "vertex_ai/claude-haiku-4-5@20251001": { - "deprecation_date": "2026-10-15", "cache_creation_input_token_cost": 1.25e-06, "cache_creation_input_token_cost_above_1hr": 2e-06, "cache_read_input_token_cost": 1e-07, + "deprecation_date": "2026-10-15", "input_cost_per_token": 1e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 200000, @@ -40305,18 +40306,18 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, + "prompt_cache_min_tokens": 4096, "regional_endpoint_uplift_multiplier": 1.1, "source": "https://cloud.google.com/vertex-ai/generative-ai/docs/partner-models/claude/haiku-4-5", "supports_assistant_prefill": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_native_streaming": true, - "supports_vision": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "vertex_ai/claude-3-5-sonnet": { "input_cost_per_token": 3e-06, @@ -40448,10 +40449,10 @@ "supports_vision": true }, "vertex_ai/claude-opus-4": { - "deprecation_date": "2026-05-14", "cache_creation_input_token_cost": 1.875e-05, "cache_creation_input_token_cost_above_1hr": 3e-05, "cache_read_input_token_cost": 1.5e-06, + "deprecation_date": "2026-05-14", "input_cost_per_token": 1.5e-05, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 200000, @@ -40459,6 +40460,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -40472,14 +40474,13 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "vertex_ai/claude-opus-4-1": { - "deprecation_date": "2026-08-05", "cache_creation_input_token_cost": 1.875e-05, "cache_creation_input_token_cost_above_1hr": 3e-05, "cache_read_input_token_cost": 1.5e-06, + "deprecation_date": "2026-08-05", "input_cost_per_token": 1.5e-05, "input_cost_per_token_batches": 7.5e-06, "litellm_provider": "vertex_ai-anthropic_models", @@ -40495,10 +40496,10 @@ "supports_vision": true }, "vertex_ai/claude-opus-4-1@20250805": { - "deprecation_date": "2026-08-05", "cache_creation_input_token_cost": 1.875e-05, "cache_creation_input_token_cost_above_1hr": 3e-05, "cache_read_input_token_cost": 1.5e-06, + "deprecation_date": "2026-08-05", "input_cost_per_token": 1.5e-05, "input_cost_per_token_batches": 7.5e-06, "litellm_provider": "vertex_ai-anthropic_models", @@ -40514,10 +40515,10 @@ "supports_vision": true }, "vertex_ai/claude-opus-4-5": { - "deprecation_date": "2026-11-24", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2026-11-24", "input_cost_per_token": 5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 200000, @@ -40525,6 +40526,7 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, @@ -40534,20 +40536,19 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "vertex_ai/claude-opus-4-5@20251101": { - "deprecation_date": "2026-11-24", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2026-11-24", "input_cost_per_token": 5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 200000, @@ -40555,6 +40556,7 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, @@ -40564,23 +40566,20 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, - "supports_pdf_input": true, - "supports_prompt_caching": true, - "supports_reasoning": true, - "supports_response_schema": true, - "supports_tool_choice": true, - "supports_vision": true, "supports_native_streaming": true, "supports_output_config": true, - "prompt_cache_min_tokens": 4096 + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true }, "vertex_ai/claude-opus-4-6": { - "deprecation_date": "2027-02-05", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-02-05", "input_cost_per_token": 5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -40588,31 +40587,32 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "vertex_ai/claude-opus-4-6@default": { - "deprecation_date": "2027-02-05", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-02-05", "input_cost_per_token": 5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -40620,31 +40620,32 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "vertex_ai/claude-opus-4-7": { - "deprecation_date": "2027-04-16", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-04-16", "input_cost_per_token": 5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -40652,14 +40653,18 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 2048, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -40667,17 +40672,13 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "vertex_ai/claude-opus-4-7@default": { - "deprecation_date": "2027-04-16", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-04-16", "input_cost_per_token": 5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -40685,14 +40686,18 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 2048, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -40700,17 +40705,13 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "vertex_ai/claude-fable-5": { - "deprecation_date": "2027-06-08", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 1.25e-05, "cache_creation_input_token_cost_above_1hr": 2e-05, "cache_read_input_token_cost": 1e-06, + "deprecation_date": "2027-06-08", "input_cost_per_token": 1e-05, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -40718,16 +40719,19 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 5e-05, + "prompt_cache_min_tokens": 512, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "thinking_always_on": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -40736,16 +40740,13 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 512 + "thinking_always_on": true }, "vertex_ai/claude-fable-5@default": { - "deprecation_date": "2027-06-08", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 1.25e-05, "cache_creation_input_token_cost_above_1hr": 2e-05, "cache_read_input_token_cost": 1e-06, + "deprecation_date": "2027-06-08", "input_cost_per_token": 1e-05, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -40753,16 +40754,19 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 5e-05, + "prompt_cache_min_tokens": 512, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "thinking_always_on": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -40771,17 +40775,13 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 512 + "thinking_always_on": true }, "vertex_ai/claude-opus-5": { - "deprecation_date": "2027-01-24", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_mid_conversation_system": true, - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-01-24", "input_cost_per_token": 5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -40789,14 +40789,19 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 512, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -40804,18 +40809,13 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 512 + "supports_xhigh_reasoning_effort": true }, "vertex_ai/claude-opus-5@default": { - "deprecation_date": "2027-01-24", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_mid_conversation_system": true, - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-01-24", "input_cost_per_token": 5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -40823,14 +40823,19 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 512, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -40838,18 +40843,13 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 512 + "supports_xhigh_reasoning_effort": true }, "vertex_ai/claude-opus-4-8": { - "deprecation_date": "2027-05-28", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_mid_conversation_system": true, - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-05-28", "input_cost_per_token": 5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -40857,14 +40857,19 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 1024, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -40872,18 +40877,13 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "vertex_ai/claude-opus-4-8@default": { - "deprecation_date": "2027-05-28", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_mid_conversation_system": true, - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-05-28", "input_cost_per_token": 5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -40891,14 +40891,19 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 1024, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -40906,20 +40911,17 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "vertex_ai/claude-sonnet-4-5": { - "deprecation_date": "2026-09-29", "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "deprecation_date": "2026-09-29", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "input_cost_per_token_batches": 1.5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 200000, @@ -40927,7 +40929,9 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, "output_cost_per_token_batches": 7.5e-06, + "prompt_cache_min_tokens": 1024, "regional_endpoint_uplift_multiplier": 1.1, "supports_assistant_prefill": true, "supports_computer_use": true, @@ -40937,16 +40941,13 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "vertex_ai/claude-sonnet-5": { - "deprecation_date": "2026-12-24", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 2.5e-06, "cache_creation_input_token_cost_above_1hr": 4e-06, "cache_read_input_token_cost": 2e-07, + "deprecation_date": "2026-12-24", "input_cost_per_token": 2e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -40954,6 +40955,8 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1e-05, + "prompt_cache_min_tokens": 1024, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -40963,6 +40966,8 @@ "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -40970,13 +40975,9 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "vertex_ai/claude-sonnet-4-6": { - "regional_endpoint_uplift_multiplier": 1.1, - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, "cache_read_input_token_cost": 3e-07, @@ -40987,34 +40988,36 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1.5e-05, - "supports_assistant_prefill": true, - "supports_computer_use": true, - "supports_function_calling": true, - "supports_pdf_input": true, - "supports_prompt_caching": true, - "supports_reasoning": true, - "supports_response_schema": true, - "supports_max_reasoning_effort": true, - "supports_tool_choice": true, - "supports_vision": true, + "prompt_cache_min_tokens": 1024, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, + "supports_assistant_prefill": true, + "supports_computer_use": true, + "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, "supports_output_config": true, - "prompt_cache_min_tokens": 1024 + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true }, "vertex_ai/claude-sonnet-4-5@20250929": { - "deprecation_date": "2026-09-29", "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "deprecation_date": "2026-09-29", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "input_cost_per_token_batches": 1.5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 200000, @@ -41022,25 +41025,26 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, "output_cost_per_token_batches": 7.5e-06, + "prompt_cache_min_tokens": 1024, "regional_endpoint_uplift_multiplier": 1.1, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_streaming": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "vertex_ai/claude-opus-4@20250514": { - "deprecation_date": "2026-05-14", "cache_creation_input_token_cost": 1.875e-05, "cache_creation_input_token_cost_above_1hr": 3e-05, "cache_read_input_token_cost": 1.5e-06, + "deprecation_date": "2026-05-14", "input_cost_per_token": 1.5e-05, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 200000, @@ -41048,6 +41052,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -41061,25 +41066,25 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "vertex_ai/claude-sonnet-4": { - "deprecation_date": "2026-05-14", "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "deprecation_date": "2026-05-14", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -41093,25 +41098,25 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "vertex_ai/claude-sonnet-4@20250514": { - "deprecation_date": "2026-05-14", "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "deprecation_date": "2026-05-14", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -41125,8 +41130,7 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "vertex_ai/mistralai/codestral-2@001": { "input_cost_per_token": 3e-07, @@ -41262,8 +41266,8 @@ "supports_tool_choice": true }, "vertex_ai/gemini-2.5-flash-image": { - "deprecation_date": "2026-10-02", "cache_read_input_token_cost": 3e-08, + "deprecation_date": "2026-10-02", "input_cost_per_audio_token": 1e-06, "input_cost_per_token": 3e-07, "litellm_provider": "vertex_ai-language-models", @@ -41294,6 +41298,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -41304,8 +41309,7 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": false, - "tpm": 8000000, - "supports_image_size": false + "tpm": 8000000 }, "vertex_ai/gemini-3-pro-image": { "deprecation_date": "2027-05-28", @@ -41321,8 +41325,8 @@ "output_cost_per_image_token": 0.00012, "output_cost_per_token": 1.2e-05, "output_cost_per_token_batches": 6e-06, - "supports_reasoning": false, - "source": "https://docs.cloud.google.com/vertex-ai/generative-ai/docs/models/gemini/3-pro-image" + "source": "https://docs.cloud.google.com/vertex-ai/generative-ai/docs/models/gemini/3-pro-image", + "supports_reasoning": false }, "vertex_ai/gemini-3-pro-image-preview": { "input_cost_per_image": 0.0011, @@ -41337,8 +41341,8 @@ "output_cost_per_image_token": 0.00012, "output_cost_per_token": 1.2e-05, "output_cost_per_token_batches": 6e-06, - "supports_reasoning": false, - "source": "https://docs.cloud.google.com/vertex-ai/generative-ai/docs/models/gemini/3-pro-image" + "source": "https://docs.cloud.google.com/vertex-ai/generative-ai/docs/models/gemini/3-pro-image", + "supports_reasoning": false }, "vertex_ai/gemini-3.1-flash-image": { "deprecation_date": "2027-05-28", @@ -41352,8 +41356,8 @@ "output_cost_per_image": 0.0672, "output_cost_per_image_token": 6e-05, "output_cost_per_token": 3e-06, - "supports_reasoning": false, - "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models" + "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models", + "supports_reasoning": false }, "vertex_ai/gemini-3.1-flash-image-preview": { "input_cost_per_image": 0.00056, @@ -41366,8 +41370,8 @@ "output_cost_per_image": 0.0672, "output_cost_per_image_token": 6e-05, "output_cost_per_token": 3e-06, - "supports_reasoning": false, - "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models" + "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models", + "supports_reasoning": false }, "vertex_ai/gemini-3.1-flash-lite-image": { "cache_read_input_token_cost": 2.5e-08, @@ -41418,6 +41422,11 @@ "mode": "chat", "output_cost_per_reasoning_token": 1.5e-06, "output_cost_per_token": 1.5e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models", "supported_endpoints": [ "/v1/chat/completions", @@ -41436,6 +41445,7 @@ "supports_audio_input": true, "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -41447,28 +41457,27 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "vertex_ai/gemini-3.1-flash-lite": { - "deprecation_date": "2027-05-07", "cache_read_input_token_cost": 2.5e-08, "cache_read_input_token_cost_flex": 1.25e-08, + "cache_read_input_token_cost_per_audio_token": 5e-08, "cache_read_input_token_cost_priority": 4.5e-08, + "deprecation_date": "2027-05-07", "input_cost_per_audio_token": 5e-07, "input_cost_per_token": 2.5e-07, "input_cost_per_token_batches": 1.25e-07, "input_cost_per_token_flex": 1.25e-07, "input_cost_per_token_priority": 4.5e-07, "litellm_provider": "vertex_ai-language-models", + "max_audio_length_hours": 8.4, + "max_audio_per_prompt": 1, + "max_images_per_prompt": 3000, "max_input_tokens": 1048576, "max_output_tokens": 65536, "max_tokens": 65536, + "max_videos_per_prompt": 10, "mode": "chat", "output_cost_per_reasoning_token": 1.5e-06, "output_cost_per_token": 1.5e-06, @@ -41476,6 +41485,11 @@ "output_cost_per_token_flex": 7.5e-07, "output_cost_per_token_priority": 2.7e-06, "regional_endpoint_uplift_multiplier": 1.1, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models", "supported_endpoints": [ "/v1/chat/completions", @@ -41493,39 +41507,29 @@ ], "supports_audio_input": true, "supports_audio_output": false, + "supports_code_execution": true, + "supports_file_search": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_tool_choice": true, "supports_url_context": true, "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, - "web_search_billing_unit": "per_query", - "cache_read_input_token_cost_per_audio_token": 5e-08, - "max_audio_length_hours": 8.4, - "max_audio_per_prompt": 1, - "max_images_per_prompt": 3000, - "max_videos_per_prompt": 10, - "supports_code_execution": true, - "supports_file_search": true, - "supports_service_tier": true + "web_search_billing_unit": "per_query" }, "vertex_ai/gemini-3.5-flash-lite": { - "deprecation_date": "2027-07-21", "cache_read_input_token_cost": 3e-08, "cache_read_input_token_cost_flex": 2e-08, "cache_read_input_token_cost_priority": 5e-08, + "deprecation_date": "2027-07-21", "input_cost_per_token": 3e-07, "input_cost_per_token_batches": 1.5e-07, "input_cost_per_token_flex": 1.5e-07, @@ -41541,6 +41545,11 @@ "output_cost_per_token_flex": 1.25e-06, "output_cost_per_token_priority": 4.5e-06, "regional_endpoint_uplift_multiplier": 1.1, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models", "supported_endpoints": [ "/v1/chat/completions", @@ -41559,6 +41568,7 @@ "supports_audio_input": true, "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -41570,12 +41580,6 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "vertex_ai/deep-research-pro-preview-12-2025": { @@ -42059,17 +42063,17 @@ "litellm_provider": "vertex_ai", "mode": "ocr", "ocr_cost_per_page": 0.0005, + "source": "https://cloud.google.com/generative-ai-app-builder/pricing", "supported_endpoints": [ "/v1/ocr" - ], - "source": "https://cloud.google.com/generative-ai-app-builder/pricing" + ] }, "vertex_ai/deepseek-ai/deepseek-ocr-maas": { + "input_cost_per_token": 3e-07, "litellm_provider": "vertex_ai", "mode": "ocr", - "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, "ocr_cost_per_page": 0.0003, + "output_cost_per_token": 1.2e-06, "source": "https://cloud.google.com/vertex-ai/pricing", "supported_regions": [ "us-central1" @@ -42500,55 +42504,55 @@ "output_cost_per_token": 0.0 }, "wandb/openai/gpt-oss-120b": { - "max_tokens": 131000, + "input_cost_per_token": 3e-08, + "litellm_provider": "wandb", "max_input_tokens": 131000, "max_output_tokens": 131000, - "input_cost_per_token": 3e-08, - "output_cost_per_token": 1.7e-07, - "litellm_provider": "wandb", + "max_tokens": 131000, "mode": "chat", + "output_cost_per_token": 1.7e-07, "source": "https://wandb.ai/site/pricing/tokens/" }, "wandb/openai/gpt-oss-20b": { - "max_tokens": 131000, + "input_cost_per_token": 3e-08, + "litellm_provider": "wandb", "max_input_tokens": 131000, "max_output_tokens": 131000, - "input_cost_per_token": 3e-08, - "output_cost_per_token": 1.3e-07, - "litellm_provider": "wandb", + "max_tokens": 131000, "mode": "chat", + "output_cost_per_token": 1.3e-07, "source": "https://wandb.ai/site/pricing/tokens/" }, "wandb/Qwen/Qwen3-Coder-480B-A35B-Instruct": { - "max_tokens": 262000, + "deprecation_date": "2026-08-25", + "input_cost_per_token": 1e-06, + "litellm_provider": "wandb", "max_input_tokens": 262000, "max_output_tokens": 262000, - "input_cost_per_token": 1e-06, - "output_cost_per_token": 1.5e-06, - "litellm_provider": "wandb", + "max_tokens": 262000, "mode": "chat", - "deprecation_date": "2026-08-25", + "output_cost_per_token": 1.5e-06, "source": "https://wandb.ai/site/pricing/tokens/" }, "wandb/moonshotai/Kimi-K2-Instruct": { - "max_tokens": 128000, + "deprecation_date": "2026-03-04", + "input_cost_per_token": 6e-07, + "litellm_provider": "wandb", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 6e-07, - "output_cost_per_token": 2.5e-06, - "litellm_provider": "wandb", + "max_tokens": 128000, "mode": "chat", - "deprecation_date": "2026-03-04" + "output_cost_per_token": 2.5e-06 }, "wandb/moonshotai/Kimi-K2.5": { - "max_tokens": 262144, - "max_input_tokens": 262144, - "max_output_tokens": 262144, "cache_read_input_token_cost": 1e-07, "input_cost_per_token": 6e-07, - "output_cost_per_token": 3e-06, "litellm_provider": "wandb", + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "max_tokens": 262144, "mode": "chat", + "output_cost_per_token": 3e-06, "source": "https://wandb.ai/inference/coreweave/cw_moonshotai_Kimi-K2.5", "supports_function_calling": true, "supports_reasoning": true, @@ -42556,47 +42560,47 @@ "supports_vision": true }, "wandb/MiniMaxAI/MiniMax-M2.5": { - "max_tokens": 197000, + "deprecation_date": "2026-08-25", + "input_cost_per_token": 3e-07, + "litellm_provider": "wandb", "max_input_tokens": 197000, "max_output_tokens": 197000, - "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "wandb", + "max_tokens": 197000, "mode": "chat", + "output_cost_per_token": 1.2e-06, "source": "https://wandb.ai/site/pricing/tokens/", "supports_function_calling": true, "supports_reasoning": true, - "supports_response_schema": true, - "deprecation_date": "2026-08-25" + "supports_response_schema": true }, "wandb/meta-llama/Llama-3.1-8B-Instruct": { - "max_tokens": 128000, + "input_cost_per_token": 2.2e-07, + "litellm_provider": "wandb", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 2.2e-07, - "output_cost_per_token": 2.2e-07, - "litellm_provider": "wandb", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 2.2e-07, "source": "https://wandb.ai/site/pricing/tokens/" }, "wandb/deepseek-ai/DeepSeek-V3.1": { - "max_tokens": 161000, + "input_cost_per_token": 5.5e-07, + "litellm_provider": "wandb", "max_input_tokens": 161000, "max_output_tokens": 161000, - "input_cost_per_token": 5.5e-07, - "output_cost_per_token": 1.65e-06, - "litellm_provider": "wandb", + "max_tokens": 161000, "mode": "chat", + "output_cost_per_token": 1.65e-06, "source": "https://wandb.ai/site/pricing/tokens/" }, "wandb/meta-llama/Llama-3.3-70B-Instruct": { - "max_tokens": 128000, + "input_cost_per_token": 7.1e-07, + "litellm_provider": "wandb", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 7.1e-07, - "output_cost_per_token": 7.1e-07, - "litellm_provider": "wandb", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 7.1e-07, "source": "https://wandb.ai/site/pricing/tokens/" }, "watsonx/ibm/granite-3-8b-instruct": { @@ -42636,322 +42640,322 @@ "supports_vision": false }, "watsonx/bigscience/mt0-xxl-13b": { - "max_tokens": 8192, + "input_cost_per_token": 0.0005, + "litellm_provider": "watsonx", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 0.0005, - "output_cost_per_token": 0.002, - "litellm_provider": "watsonx", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 0.002, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/core42/jais-13b-chat": { - "max_tokens": 8192, + "input_cost_per_token": 0.0005, + "litellm_provider": "watsonx", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 0.0005, - "output_cost_per_token": 0.002, - "litellm_provider": "watsonx", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 0.002, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/google/flan-t5-xl-3b": { - "max_tokens": 8192, + "input_cost_per_token": 6e-07, + "litellm_provider": "watsonx", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 6e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "watsonx", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 6e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/ibm/granite-13b-chat-v2": { - "max_tokens": 8192, + "input_cost_per_token": 6e-07, + "litellm_provider": "watsonx", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 6e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "watsonx", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 6e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/ibm/granite-13b-instruct-v2": { - "max_tokens": 8192, + "input_cost_per_token": 6e-07, + "litellm_provider": "watsonx", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 6e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "watsonx", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 6e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/ibm/granite-3-3-8b-instruct": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "watsonx", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "watsonx", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 2e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": false }, "watsonx/ibm/granite-4-h-small": { - "max_tokens": 20480, + "input_cost_per_token": 6e-08, + "litellm_provider": "watsonx", "max_input_tokens": 131072, "max_output_tokens": 20480, - "input_cost_per_token": 6e-08, - "output_cost_per_token": 2.5e-07, - "litellm_provider": "watsonx", + "max_tokens": 20480, "mode": "chat", + "output_cost_per_token": 2.5e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": false }, "watsonx/ibm/granite-guardian-3-2-2b": { - "max_tokens": 8192, + "input_cost_per_token": 1e-07, + "litellm_provider": "watsonx", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "watsonx", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/ibm/granite-guardian-3-3-8b": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "watsonx", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "watsonx", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 2e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/ibm/granite-ttm-1024-96-r2": { - "max_tokens": 512, + "input_cost_per_token": 3.8e-07, + "litellm_provider": "watsonx", "max_input_tokens": 512, "max_output_tokens": 512, - "input_cost_per_token": 3.8e-07, - "output_cost_per_token": 3.8e-07, - "litellm_provider": "watsonx", + "max_tokens": 512, "mode": "chat", + "output_cost_per_token": 3.8e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/ibm/granite-ttm-1536-96-r2": { - "max_tokens": 512, + "input_cost_per_token": 3.8e-07, + "litellm_provider": "watsonx", "max_input_tokens": 512, "max_output_tokens": 512, - "input_cost_per_token": 3.8e-07, - "output_cost_per_token": 3.8e-07, - "litellm_provider": "watsonx", + "max_tokens": 512, "mode": "chat", + "output_cost_per_token": 3.8e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/ibm/granite-ttm-512-96-r2": { - "max_tokens": 512, + "input_cost_per_token": 3.8e-07, + "litellm_provider": "watsonx", "max_input_tokens": 512, "max_output_tokens": 512, - "input_cost_per_token": 3.8e-07, - "output_cost_per_token": 3.8e-07, - "litellm_provider": "watsonx", + "max_tokens": 512, "mode": "chat", + "output_cost_per_token": 3.8e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/ibm/granite-vision-3-2-2b": { - "max_tokens": 8192, + "input_cost_per_token": 1e-07, + "litellm_provider": "watsonx", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "watsonx", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": true }, "watsonx/meta-llama/llama-3-2-11b-vision-instruct": { - "max_tokens": 128000, + "input_cost_per_token": 3.5e-07, + "litellm_provider": "watsonx", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 3.5e-07, - "output_cost_per_token": 3.5e-07, - "litellm_provider": "watsonx", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.5e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": true }, "watsonx/meta-llama/llama-3-2-1b-instruct": { - "max_tokens": 128000, + "input_cost_per_token": 1e-07, + "litellm_provider": "watsonx", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "watsonx", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": false }, "watsonx/meta-llama/llama-3-2-3b-instruct": { - "max_tokens": 128000, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "watsonx", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 1.5e-07, - "litellm_provider": "watsonx", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1.5e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": false }, "watsonx/meta-llama/llama-3-2-90b-vision-instruct": { - "max_tokens": 128000, + "input_cost_per_token": 2e-06, + "litellm_provider": "watsonx", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 2e-06, - "output_cost_per_token": 2e-06, - "litellm_provider": "watsonx", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 2e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": true }, "watsonx/meta-llama/llama-3-3-70b-instruct": { - "max_tokens": 128000, + "input_cost_per_token": 7.1e-07, + "litellm_provider": "watsonx", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 7.1e-07, - "output_cost_per_token": 7.1e-07, - "litellm_provider": "watsonx", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 7.1e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": false }, "watsonx/meta-llama/llama-4-maverick-17b": { - "max_tokens": 128000, + "input_cost_per_token": 3.5e-07, + "litellm_provider": "watsonx", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 3.5e-07, - "output_cost_per_token": 1.4e-06, - "litellm_provider": "watsonx", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1.4e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": false }, "watsonx/meta-llama/llama-guard-3-11b-vision": { - "max_tokens": 128000, + "input_cost_per_token": 3.5e-07, + "litellm_provider": "watsonx", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 3.5e-07, - "output_cost_per_token": 3.5e-07, - "litellm_provider": "watsonx", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.5e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": true }, "watsonx/mistralai/mistral-medium-2505": { - "max_tokens": 128000, + "input_cost_per_token": 3e-06, + "litellm_provider": "watsonx", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 3e-06, - "output_cost_per_token": 1e-05, - "litellm_provider": "watsonx", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1e-05, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": false }, "watsonx/mistralai/mistral-small-2503": { - "max_tokens": 32000, + "input_cost_per_token": 1e-07, + "litellm_provider": "watsonx", "max_input_tokens": 32000, "max_output_tokens": 32000, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 3e-07, - "litellm_provider": "watsonx", + "max_tokens": 32000, "mode": "chat", + "output_cost_per_token": 3e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": false }, "watsonx/mistralai/mistral-small-3-1-24b-instruct-2503": { - "max_tokens": 32000, + "input_cost_per_token": 1e-07, + "litellm_provider": "watsonx", "max_input_tokens": 32000, "max_output_tokens": 32000, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 3e-07, - "litellm_provider": "watsonx", + "max_tokens": 32000, "mode": "chat", + "output_cost_per_token": 3e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": false }, "watsonx/mistralai/pixtral-12b-2409": { - "max_tokens": 128000, + "input_cost_per_token": 3.5e-07, + "litellm_provider": "watsonx", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 3.5e-07, - "output_cost_per_token": 3.5e-07, - "litellm_provider": "watsonx", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.5e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": true }, "watsonx/openai/gpt-oss-120b": { - "max_tokens": 8192, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "watsonx", "max_input_tokens": 131072, "max_output_tokens": 8192, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "watsonx", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 6e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/sdaia/allam-1-13b-instruct": { - "max_tokens": 8192, + "input_cost_per_token": 1.8e-06, + "litellm_provider": "watsonx", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 1.8e-06, - "output_cost_per_token": 1.8e-06, - "litellm_provider": "watsonx", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.8e-06, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/whisper-large-v3-turbo": { "input_cost_per_second": 0.0001, - "output_cost_per_second": 0.0001, "litellm_provider": "watsonx", "mode": "audio_transcription", + "output_cost_per_second": 0.0001, "supported_endpoints": [ "/v1/audio/transcriptions" ] @@ -43046,6 +43050,7 @@ }, "xai/grok-3": { "cache_read_input_token_cost": 7.5e-07, + "deprecation_date": "2026-05-15", "input_cost_per_token": 3e-06, "litellm_provider": "xai", "max_input_tokens": 131072, @@ -43058,8 +43063,7 @@ "supports_prompt_caching": true, "supports_response_schema": false, "supports_tool_choice": true, - "supports_web_search": true, - "deprecation_date": "2026-05-15" + "supports_web_search": true }, "xai/grok-3-beta": { "cache_read_input_token_cost": 7.5e-07, @@ -43111,6 +43115,7 @@ }, "xai/grok-3-latest": { "cache_read_input_token_cost": 7.5e-07, + "deprecation_date": "2026-05-15", "input_cost_per_token": 3e-06, "litellm_provider": "xai", "max_input_tokens": 131072, @@ -43123,8 +43128,7 @@ "supports_prompt_caching": true, "supports_response_schema": false, "supports_tool_choice": true, - "supports_web_search": true, - "deprecation_date": "2026-05-15" + "supports_web_search": true }, "xai/grok-3-mini": { "cache_read_input_token_cost": 7.5e-08, @@ -43231,6 +43235,7 @@ "supports_web_search": true }, "xai/grok-4": { + "deprecation_date": "2026-05-15", "input_cost_per_token": 3e-06, "litellm_provider": "xai", "max_input_tokens": 256000, @@ -43242,11 +43247,11 @@ "supports_function_calling": true, "supports_prompt_caching": true, "supports_tool_choice": true, - "supports_web_search": true, - "deprecation_date": "2026-05-15" + "supports_web_search": true }, "xai/grok-4-fast-reasoning": { "cache_read_input_token_cost": 5e-08, + "deprecation_date": "2026-05-15", "input_cost_per_token": 2e-07, "input_cost_per_token_above_128k_tokens": 4e-07, "litellm_provider": "xai", @@ -43260,11 +43265,11 @@ "supports_function_calling": true, "supports_prompt_caching": true, "supports_tool_choice": true, - "supports_web_search": true, - "deprecation_date": "2026-05-15" + "supports_web_search": true }, "xai/grok-4-fast-non-reasoning": { "cache_read_input_token_cost": 5e-08, + "deprecation_date": "2026-05-15", "input_cost_per_token": 2e-07, "input_cost_per_token_above_128k_tokens": 4e-07, "litellm_provider": "xai", @@ -43278,10 +43283,10 @@ "supports_function_calling": true, "supports_prompt_caching": true, "supports_tool_choice": true, - "supports_web_search": true, - "deprecation_date": "2026-05-15" + "supports_web_search": true }, "xai/grok-4-0709": { + "deprecation_date": "2026-05-15", "input_cost_per_token": 3e-06, "input_cost_per_token_above_128k_tokens": 6e-06, "litellm_provider": "xai", @@ -43295,10 +43300,10 @@ "supports_function_calling": true, "supports_prompt_caching": true, "supports_tool_choice": true, - "supports_web_search": true, - "deprecation_date": "2026-05-15" + "supports_web_search": true }, "xai/grok-4-latest": { + "deprecation_date": "2026-05-15", "input_cost_per_token": 3e-06, "input_cost_per_token_above_128k_tokens": 6e-06, "litellm_provider": "xai", @@ -43312,8 +43317,7 @@ "supports_function_calling": true, "supports_prompt_caching": true, "supports_tool_choice": true, - "supports_web_search": true, - "deprecation_date": "2026-05-15" + "supports_web_search": true }, "xai/grok-4-1-fast": { "cache_read_input_token_cost": 5e-08, @@ -43338,6 +43342,7 @@ }, "xai/grok-4-1-fast-reasoning": { "cache_read_input_token_cost": 5e-08, + "deprecation_date": "2026-05-15", "input_cost_per_token": 2e-07, "input_cost_per_token_above_128k_tokens": 4e-07, "litellm_provider": "xai", @@ -43355,11 +43360,11 @@ "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "deprecation_date": "2026-05-15" + "supports_web_search": true }, "xai/grok-4-1-fast-reasoning-latest": { "cache_read_input_token_cost": 5e-08, + "deprecation_date": "2026-05-15", "input_cost_per_token": 2e-07, "input_cost_per_token_above_128k_tokens": 4e-07, "litellm_provider": "xai", @@ -43377,11 +43382,11 @@ "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "deprecation_date": "2026-05-15" + "supports_web_search": true }, "xai/grok-4-1-fast-non-reasoning": { "cache_read_input_token_cost": 5e-08, + "deprecation_date": "2026-05-15", "input_cost_per_token": 2e-07, "input_cost_per_token_above_128k_tokens": 4e-07, "litellm_provider": "xai", @@ -43398,11 +43403,11 @@ "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "deprecation_date": "2026-05-15" + "supports_web_search": true }, "xai/grok-4-1-fast-non-reasoning-latest": { "cache_read_input_token_cost": 5e-08, + "deprecation_date": "2026-05-15", "input_cost_per_token": 2e-07, "input_cost_per_token_above_128k_tokens": 4e-07, "litellm_provider": "xai", @@ -43419,91 +43424,90 @@ "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "deprecation_date": "2026-05-15" + "supports_web_search": true }, "xai/grok-4.20-multi-agent-beta-0309": { "cache_read_input_token_cost": 2e-07, + "cache_read_input_token_cost_above_200k_tokens": 4e-07, "input_cost_per_token": 1.25e-06, + "input_cost_per_token_above_200k_tokens": 2.5e-06, "litellm_provider": "xai", "max_input_tokens": 1000000, "max_output_tokens": 1000000, "max_tokens": 1000000, "mode": "chat", "output_cost_per_token": 2.5e-06, + "output_cost_per_token_above_200k_tokens": 5e-06, "source": "https://docs.x.ai/docs/models", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, + "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "input_cost_per_token_above_200k_tokens": 2.5e-06, - "output_cost_per_token_above_200k_tokens": 5e-06, - "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "supports_response_schema": true + "supports_web_search": true }, "xai/grok-4.20-beta-0309-reasoning": { "cache_read_input_token_cost": 2e-07, + "cache_read_input_token_cost_above_200k_tokens": 4e-07, "input_cost_per_token": 1.25e-06, + "input_cost_per_token_above_200k_tokens": 2.5e-06, "litellm_provider": "xai", "max_input_tokens": 1000000, "max_output_tokens": 1000000, "max_tokens": 1000000, "mode": "chat", "output_cost_per_token": 2.5e-06, + "output_cost_per_token_above_200k_tokens": 5e-06, "source": "https://docs.x.ai/docs/models", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, + "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "input_cost_per_token_above_200k_tokens": 2.5e-06, - "output_cost_per_token_above_200k_tokens": 5e-06, - "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "supports_response_schema": true + "supports_web_search": true }, "xai/grok-4.20-0309-reasoning": { "cache_read_input_token_cost": 2e-07, + "cache_read_input_token_cost_above_200k_tokens": 4e-07, "input_cost_per_token": 1.25e-06, + "input_cost_per_token_above_200k_tokens": 2.5e-06, "litellm_provider": "xai", "max_input_tokens": 1000000, "max_output_tokens": 1000000, "max_tokens": 1000000, "mode": "chat", "output_cost_per_token": 2.5e-06, + "output_cost_per_token_above_200k_tokens": 5e-06, "source": "https://docs.x.ai/docs/models", "supports_function_calling": true, + "supports_prompt_caching": true, "supports_reasoning": true, + "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "input_cost_per_token_above_200k_tokens": 2.5e-06, - "output_cost_per_token_above_200k_tokens": 5e-06, - "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "supports_prompt_caching": true, - "supports_response_schema": true + "supports_web_search": true }, "xai/grok-4.20-beta-0309-non-reasoning": { "cache_read_input_token_cost": 2e-07, + "cache_read_input_token_cost_above_200k_tokens": 4e-07, "input_cost_per_token": 1.25e-06, + "input_cost_per_token_above_200k_tokens": 2.5e-06, "litellm_provider": "xai", "max_input_tokens": 1000000, "max_output_tokens": 1000000, "max_tokens": 1000000, "mode": "chat", "output_cost_per_token": 2.5e-06, + "output_cost_per_token_above_200k_tokens": 5e-06, "source": "https://docs.x.ai/docs/models", "supports_function_calling": true, "supports_prompt_caching": true, + "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "input_cost_per_token_above_200k_tokens": 2.5e-06, - "output_cost_per_token_above_200k_tokens": 5e-06, - "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "supports_response_schema": true + "supports_web_search": true }, "xai/grok-4.3": { "cache_read_input_token_cost": 2e-07, @@ -43625,66 +43629,66 @@ }, "xai/grok-code-fast": { "cache_read_input_token_cost": 2e-07, + "cache_read_input_token_cost_above_200k_tokens": 4e-07, + "deprecation_date": "2026-05-15", "input_cost_per_token": 1e-06, + "input_cost_per_token_above_200k_tokens": 2e-06, "litellm_provider": "xai", "max_input_tokens": 256000, "max_output_tokens": 256000, "max_tokens": 256000, "mode": "chat", "output_cost_per_token": 2e-06, + "output_cost_per_token_above_200k_tokens": 4e-06, "source": "https://docs.x.ai/docs/models", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "input_cost_per_token_above_200k_tokens": 2e-06, - "output_cost_per_token_above_200k_tokens": 4e-06, - "cache_read_input_token_cost_above_200k_tokens": 4e-07, "supports_response_schema": true, - "supports_vision": true, - "deprecation_date": "2026-05-15" + "supports_tool_choice": true, + "supports_vision": true }, "xai/grok-code-fast-1": { "cache_read_input_token_cost": 2e-07, + "cache_read_input_token_cost_above_200k_tokens": 4e-07, + "deprecation_date": "2026-05-15", "input_cost_per_token": 1e-06, + "input_cost_per_token_above_200k_tokens": 2e-06, "litellm_provider": "xai", "max_input_tokens": 256000, "max_output_tokens": 256000, "max_tokens": 256000, "mode": "chat", "output_cost_per_token": 2e-06, + "output_cost_per_token_above_200k_tokens": 4e-06, "source": "https://docs.x.ai/docs/models", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "input_cost_per_token_above_200k_tokens": 2e-06, - "output_cost_per_token_above_200k_tokens": 4e-06, - "cache_read_input_token_cost_above_200k_tokens": 4e-07, "supports_response_schema": true, - "supports_vision": true, - "deprecation_date": "2026-05-15" + "supports_tool_choice": true, + "supports_vision": true }, "xai/grok-code-fast-1-0825": { "cache_read_input_token_cost": 2e-07, + "cache_read_input_token_cost_above_200k_tokens": 4e-07, + "deprecation_date": "2026-05-15", "input_cost_per_token": 1e-06, + "input_cost_per_token_above_200k_tokens": 2e-06, "litellm_provider": "xai", "max_input_tokens": 256000, "max_output_tokens": 256000, "max_tokens": 256000, "mode": "chat", "output_cost_per_token": 2e-06, + "output_cost_per_token_above_200k_tokens": 4e-06, "source": "https://docs.x.ai/docs/models", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "input_cost_per_token_above_200k_tokens": 2e-06, - "output_cost_per_token_above_200k_tokens": 4e-06, - "cache_read_input_token_cost_above_200k_tokens": 4e-07, "supports_response_schema": true, - "supports_vision": true, - "deprecation_date": "2026-05-15" + "supports_tool_choice": true, + "supports_vision": true }, "xai/grok-vision-beta": { "input_cost_per_image": 5e-06, @@ -43708,26 +43712,26 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "zai.glm-5": { "input_cost_per_token": 1e-06, - "output_cost_per_token": 3.2e-06, "litellm_provider": "bedrock_converse", "max_input_tokens": 200000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, + "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "zai.glm-4.7-flash": { "input_cost_per_token": 7e-08, @@ -43737,179 +43741,179 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 4e-07, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "zai/glm-5": { "cache_creation_input_token_cost": 0, "cache_read_input_token_cost": 2e-07, "input_cost_per_token": 1e-06, - "output_cost_per_token": 3.2e-06, "litellm_provider": "zai", "max_input_tokens": 200000, "max_output_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.2e-06, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "zai/glm-5.1": { "cache_creation_input_token_cost": 0, "cache_read_input_token_cost": 2.6e-07, "input_cost_per_token": 1.4e-06, - "output_cost_per_token": 4.4e-06, "litellm_provider": "zai", "max_input_tokens": 200000, "max_output_tokens": 128000, "mode": "chat", + "output_cost_per_token": 4.4e-06, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "zai/glm-5-code": { "cache_creation_input_token_cost": 0, "cache_read_input_token_cost": 3e-07, "input_cost_per_token": 1.2e-06, - "output_cost_per_token": 5e-06, "litellm_provider": "zai", "max_input_tokens": 200000, "max_output_tokens": 128000, "mode": "chat", + "output_cost_per_token": 5e-06, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "zai/glm-4.7": { "cache_creation_input_token_cost": 0, "cache_read_input_token_cost": 1.1e-07, "input_cost_per_token": 6e-07, - "output_cost_per_token": 2.2e-06, "litellm_provider": "zai", "max_input_tokens": 200000, "max_output_tokens": 128000, "mode": "chat", + "output_cost_per_token": 2.2e-06, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "zai/glm-4.7-flash": { "cache_creation_input_token_cost": 0, "cache_read_input_token_cost": 0, "input_cost_per_token": 0, - "output_cost_per_token": 0, "litellm_provider": "zai", "max_input_tokens": 200000, "max_output_tokens": 128000, "mode": "chat", + "output_cost_per_token": 0, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "zai/glm-4.6": { "cache_creation_input_token_cost": 0, "cache_read_input_token_cost": 1.1e-07, "input_cost_per_token": 6e-07, - "output_cost_per_token": 2.2e-06, "litellm_provider": "zai", "max_input_tokens": 200000, "max_output_tokens": 128000, "mode": "chat", + "output_cost_per_token": 2.2e-06, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "zai/glm-4.5": { "input_cost_per_token": 6e-07, - "output_cost_per_token": 2.2e-06, "litellm_provider": "zai", "max_input_tokens": 128000, "max_output_tokens": 32000, "mode": "chat", + "output_cost_per_token": 2.2e-06, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "zai/glm-4.5v": { "input_cost_per_token": 6e-07, - "output_cost_per_token": 1.8e-06, "litellm_provider": "zai", "max_input_tokens": 128000, "max_output_tokens": 32000, "mode": "chat", + "output_cost_per_token": 1.8e-06, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, "supports_tool_choice": true, - "supports_vision": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_vision": true }, "zai/glm-4.5-x": { "input_cost_per_token": 2.2e-06, - "output_cost_per_token": 8.9e-06, "litellm_provider": "zai", "max_input_tokens": 128000, "max_output_tokens": 32000, "mode": "chat", + "output_cost_per_token": 8.9e-06, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "zai/glm-4.5-air": { "input_cost_per_token": 2e-07, - "output_cost_per_token": 1.1e-06, "litellm_provider": "zai", "max_input_tokens": 128000, "max_output_tokens": 32000, "mode": "chat", + "output_cost_per_token": 1.1e-06, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "zai/glm-4.5-airx": { "input_cost_per_token": 1.1e-06, - "output_cost_per_token": 4.5e-06, "litellm_provider": "zai", "max_input_tokens": 128000, "max_output_tokens": 32000, "mode": "chat", + "output_cost_per_token": 4.5e-06, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "zai/glm-4-32b-0414-128k": { "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, "litellm_provider": "zai", "max_input_tokens": 128000, "max_output_tokens": 32000, "mode": "chat", + "output_cost_per_token": 1e-07, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "zai/glm-4.5-flash": { "input_cost_per_token": 0, - "output_cost_per_token": 0, "litellm_provider": "zai", "max_input_tokens": 128000, "max_output_tokens": 32000, "mode": "chat", + "output_cost_per_token": 0, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "vertex_ai/search_api": { "input_cost_per_query": 0.0015, @@ -44000,6 +44004,9 @@ }, "runwayml/gen4_turbo": { "litellm_provider": "runwayml", + "metadata": { + "comment": "5 credits per second @ $0.01 per credit = $0.05 per second" + }, "mode": "video_generation", "output_cost_per_video_per_second": 0.05, "source": "https://docs.dev.runwayml.com/guides/pricing/", @@ -44009,13 +44016,13 @@ ], "supported_output_modalities": [ "video" - ], - "metadata": { - "comment": "5 credits per second @ $0.01 per credit = $0.05 per second" - } + ] }, "runwayml/gen4_aleph": { "litellm_provider": "runwayml", + "metadata": { + "comment": "15 credits per second @ $0.01 per credit = $0.15 per second" + }, "mode": "video_generation", "output_cost_per_video_per_second": 0.15, "source": "https://docs.dev.runwayml.com/guides/pricing/", @@ -44025,13 +44032,13 @@ ], "supported_output_modalities": [ "video" - ], - "metadata": { - "comment": "15 credits per second @ $0.01 per credit = $0.15 per second" - } + ] }, "runwayml/gen3a_turbo": { "litellm_provider": "runwayml", + "metadata": { + "comment": "5 credits per second @ $0.01 per credit = $0.05 per second" + }, "mode": "video_generation", "output_cost_per_video_per_second": 0.05, "source": "https://docs.dev.runwayml.com/guides/pricing/", @@ -44041,15 +44048,15 @@ ], "supported_output_modalities": [ "video" - ], - "metadata": { - "comment": "5 credits per second @ $0.01 per credit = $0.05 per second" - } + ] }, "runwayml/gen4_image": { - "litellm_provider": "runwayml", - "mode": "image_generation", "input_cost_per_image": 0.05, + "litellm_provider": "runwayml", + "metadata": { + "comment": "5 credits per 720p image or 8 credits per 1080p image @ $0.01 per credit. Using 5 credits ($0.05) as base cost" + }, + "mode": "image_generation", "output_cost_per_image": 0.05, "source": "https://docs.dev.runwayml.com/guides/pricing/", "supported_modalities": [ @@ -44058,15 +44065,15 @@ ], "supported_output_modalities": [ "image" - ], - "metadata": { - "comment": "5 credits per 720p image or 8 credits per 1080p image @ $0.01 per credit. Using 5 credits ($0.05) as base cost" - } + ] }, "runwayml/gen4_image_turbo": { - "litellm_provider": "runwayml", - "mode": "image_generation", "input_cost_per_image": 0.02, + "litellm_provider": "runwayml", + "metadata": { + "comment": "2 credits per image (any resolution) @ $0.01 per credit = $0.02 per image" + }, + "mode": "image_generation", "output_cost_per_image": 0.02, "source": "https://docs.dev.runwayml.com/guides/pricing/", "supported_modalities": [ @@ -44075,1904 +44082,1901 @@ ], "supported_output_modalities": [ "image" - ], - "metadata": { - "comment": "2 credits per image (any resolution) @ $0.01 per credit = $0.02 per image" - } + ] }, "runwayml/eleven_multilingual_v2": { - "litellm_provider": "runwayml", - "mode": "audio_speech", "input_cost_per_character": 3e-07, - "source": "https://docs.dev.runwayml.com/guides/pricing/", + "litellm_provider": "runwayml", "metadata": { "comment": "Estimated cost based on standard TTS pricing. RunwayML uses ElevenLabs models." - } + }, + "mode": "audio_speech", + "source": "https://docs.dev.runwayml.com/guides/pricing/" }, "fireworks_ai/accounts/fireworks/models/qwen3-coder-480b-a35b-instruct": { - "max_tokens": 262144, + "input_cost_per_token": 4.5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 4.5e-07, - "output_cost_per_token": 1.8e-06, - "litellm_provider": "fireworks_ai", + "max_tokens": 262144, "mode": "chat", + "output_cost_per_token": 1.8e-06, "supports_reasoning": true }, "fireworks_ai/accounts/fireworks/models/flux-kontext-pro": { - "max_tokens": 4096, + "input_cost_per_token": 4e-08, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 4e-08, - "output_cost_per_token": 4e-08, - "litellm_provider": "fireworks_ai", - "mode": "image_generation" + "max_tokens": 4096, + "mode": "image_generation", + "output_cost_per_token": 4e-08 }, "fireworks_ai/accounts/fireworks/models/SSD-1B": { - "max_tokens": 4096, + "input_cost_per_token": 1.3e-10, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1.3e-10, - "output_cost_per_token": 1.3e-10, - "litellm_provider": "fireworks_ai", - "mode": "image_generation" + "max_tokens": 4096, + "mode": "image_generation", + "output_cost_per_token": 1.3e-10 }, "fireworks_ai/accounts/fireworks/models/chronos-hermes-13b-v2": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-13b": { - "max_tokens": 16384, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-13b-instruct": { - "max_tokens": 16384, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-13b-python": { - "max_tokens": 16384, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-34b": { - "max_tokens": 16384, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-34b-instruct": { - "max_tokens": 16384, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-34b-python": { - "max_tokens": 16384, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-70b": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-70b-instruct": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-70b-python": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-7b": { - "max_tokens": 16384, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-7b-instruct": { - "max_tokens": 16384, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-7b-python": { - "max_tokens": 16384, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/code-qwen-1p5-7b": { - "max_tokens": 65536, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 65536, "max_output_tokens": 65536, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/codegemma-2b": { - "max_tokens": 8192, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/codegemma-7b": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/cogito-671b-v2-p1": { - "max_tokens": 163840, + "input_cost_per_token": 1.2e-06, + "litellm_provider": "fireworks_ai", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 1.2e-06, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 163840, + "mode": "chat", + "output_cost_per_token": 1.2e-06 }, "fireworks_ai/accounts/fireworks/models/cogito-v1-preview-llama-3b": { - "max_tokens": 131072, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/cogito-v1-preview-llama-70b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/cogito-v1-preview-llama-8b": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/cogito-v1-preview-qwen-14b": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/cogito-v1-preview-qwen-32b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/flux-kontext-max": { - "max_tokens": 4096, + "input_cost_per_token": 8e-08, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 8e-08, - "output_cost_per_token": 8e-08, - "litellm_provider": "fireworks_ai", - "mode": "image_generation" + "max_tokens": 4096, + "mode": "image_generation", + "output_cost_per_token": 8e-08 }, "fireworks_ai/accounts/fireworks/models/dbrx-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 1.2e-06, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1.2e-06, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1.2e-06 }, "fireworks_ai/accounts/fireworks/models/deepseek-coder-1b-base": { - "max_tokens": 16384, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-coder-33b-instruct": { - "max_tokens": 16384, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-coder-7b-base": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-coder-7b-base-v1p5": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-coder-7b-instruct-v1p5": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-coder-v2-lite-base": { - "max_tokens": 163840, + "input_cost_per_token": 5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 163840, + "mode": "chat", + "output_cost_per_token": 5e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-coder-v2-lite-instruct": { - "max_tokens": 163840, + "input_cost_per_token": 5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 163840, + "mode": "chat", + "output_cost_per_token": 5e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-prover-v2": { - "max_tokens": 163840, + "input_cost_per_token": 1.2e-06, + "litellm_provider": "fireworks_ai", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 1.2e-06, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 163840, + "mode": "chat", + "output_cost_per_token": 1.2e-06 }, "fireworks_ai/accounts/fireworks/models/deepseek-r1-0528-distill-qwen3-8b": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-r1-distill-llama-70b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-r1-distill-llama-8b": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-r1-distill-qwen-14b": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-r1-distill-qwen-1p5b": { - "max_tokens": 131072, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-r1-distill-qwen-32b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-r1-distill-qwen-7b": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-v2-lite-chat": { - "max_tokens": 163840, + "input_cost_per_token": 5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 163840, + "mode": "chat", + "output_cost_per_token": 5e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-v2p5": { - "max_tokens": 32768, + "input_cost_per_token": 1.2e-06, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1.2e-06, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1.2e-06 }, "fireworks_ai/accounts/fireworks/models/devstral-small-2505": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/dobby-mini-unhinged-plus-llama-3-1-8b": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/dobby-unhinged-llama-3-3-70b-new": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/dolphin-2-9-2-qwen2-72b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/dolphin-2p6-mixtral-8x7b": { - "max_tokens": 32768, + "input_cost_per_token": 5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 5e-07 }, "fireworks_ai/accounts/fireworks/models/ernie-4p5-21b-a3b-pt": { - "max_tokens": 4096, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/ernie-4p5-300b-a47b-pt": { - "max_tokens": 4096, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/fare-20b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/firefunction-v1": { - "max_tokens": 32768, + "input_cost_per_token": 5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 5e-07 }, "fireworks_ai/accounts/fireworks/models/firellava-13b": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/firesearch-ocr-v6": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/fireworks-asr-large": { - "max_tokens": 4096, + "input_cost_per_token": 0.0, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 0.0, - "output_cost_per_token": 0.0, - "litellm_provider": "fireworks_ai", - "mode": "audio_transcription" + "max_tokens": 4096, + "mode": "audio_transcription", + "output_cost_per_token": 0.0 }, "fireworks_ai/accounts/fireworks/models/fireworks-asr-v2": { - "max_tokens": 4096, + "input_cost_per_token": 0.0, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 0.0, - "output_cost_per_token": 0.0, - "litellm_provider": "fireworks_ai", - "mode": "audio_transcription" + "max_tokens": 4096, + "mode": "audio_transcription", + "output_cost_per_token": 0.0 }, "fireworks_ai/accounts/fireworks/models/flux-1-dev": { - "max_tokens": 4096, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/flux-1-dev-controlnet-union": { - "max_tokens": 4096, + "input_cost_per_token": 1e-09, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1e-09, - "output_cost_per_token": 1e-09, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1e-09 }, "fireworks_ai/accounts/fireworks/models/flux-1-dev-fp8": { - "max_tokens": 4096, + "input_cost_per_token": 5e-10, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 5e-10, - "output_cost_per_token": 5e-10, - "litellm_provider": "fireworks_ai", - "mode": "image_generation" + "max_tokens": 4096, + "mode": "image_generation", + "output_cost_per_token": 5e-10 }, "fireworks_ai/accounts/fireworks/models/flux-1-schnell": { - "max_tokens": 4096, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/flux-1-schnell-fp8": { - "max_tokens": 4096, + "input_cost_per_token": 3.5e-10, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 3.5e-10, - "output_cost_per_token": 3.5e-10, - "litellm_provider": "fireworks_ai", - "mode": "image_generation" + "max_tokens": 4096, + "mode": "image_generation", + "output_cost_per_token": 3.5e-10 }, "fireworks_ai/accounts/fireworks/models/gemma-2b-it": { - "max_tokens": 8192, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/gemma-3-27b-it": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/gemma-7b": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/gemma-7b-it": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/gemma2-9b-it": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/glm-4p5v": { - "max_tokens": 131072, + "input_cost_per_token": 1.2e-06, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1.2e-06, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "fireworks_ai", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 1.2e-06, "supports_reasoning": true }, "fireworks_ai/accounts/fireworks/models/gpt-oss-safeguard-120b": { - "max_tokens": 131072, + "input_cost_per_token": 1.2e-06, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1.2e-06, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 1.2e-06 }, "fireworks_ai/accounts/fireworks/models/gpt-oss-safeguard-20b": { - "max_tokens": 131072, + "input_cost_per_token": 5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 5e-07 }, "fireworks_ai/accounts/fireworks/models/hermes-2-pro-mistral-7b": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/internvl3-38b": { - "max_tokens": 16384, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/internvl3-78b": { - "max_tokens": 16384, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/internvl3-8b": { - "max_tokens": 16384, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/japanese-stable-diffusion-xl": { - "max_tokens": 4096, + "input_cost_per_token": 1.3e-10, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1.3e-10, - "output_cost_per_token": 1.3e-10, - "litellm_provider": "fireworks_ai", - "mode": "image_generation" + "max_tokens": 4096, + "mode": "image_generation", + "output_cost_per_token": 1.3e-10 }, "fireworks_ai/accounts/fireworks/models/kat-coder": { - "max_tokens": 262144, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/kat-dev-32b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/kat-dev-72b-exp": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/llama-guard-2-8b": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/llama-guard-3-1b": { - "max_tokens": 131072, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/llama-guard-3-8b": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v2-13b": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v2-13b-chat": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v2-70b": { - "max_tokens": 4096, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v2-70b-chat": { - "max_tokens": 2048, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 2048, "max_output_tokens": 2048, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 2048, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v2-7b": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v2-7b-chat": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v3-70b-instruct": { - "max_tokens": 8192, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v3-70b-instruct-hf": { - "max_tokens": 8192, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v3-8b": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v3-8b-instruct-hf": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v3p1-405b-instruct-long": { - "max_tokens": 4096, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v3p1-70b-instruct": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v3p1-70b-instruct-1b": { - "max_tokens": 4096, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v3p1-nemotron-70b-instruct": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v3p2-1b": { - "max_tokens": 131072, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v3p2-3b": { - "max_tokens": 131072, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v3p3-70b-instruct": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/llamaguard-7b": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/llava-yi-34b": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/minimax-m1-80k": { - "max_tokens": 4096, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/minimax-m2": { - "max_tokens": 4096, + "input_cost_per_token": 3e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1.2e-06 }, "fireworks_ai/accounts/fireworks/models/ministral-3-14b-instruct-2512": { - "max_tokens": 256000, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 256000, "max_output_tokens": 256000, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 256000, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/ministral-3-3b-instruct-2512": { - "max_tokens": 256000, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 256000, "max_output_tokens": 256000, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 256000, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/ministral-3-8b-instruct-2512": { - "max_tokens": 256000, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 256000, "max_output_tokens": 256000, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 256000, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/mistral-7b": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/mistral-7b-instruct-4k": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/mistral-7b-instruct-v0p2": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/mistral-7b-instruct-v3": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/mistral-7b-v0p2": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/mistral-large-3-fp8": { - "max_tokens": 256000, + "input_cost_per_token": 1.2e-06, + "litellm_provider": "fireworks_ai", "max_input_tokens": 256000, "max_output_tokens": 256000, - "input_cost_per_token": 1.2e-06, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 256000, + "mode": "chat", + "output_cost_per_token": 1.2e-06 }, "fireworks_ai/accounts/fireworks/models/mistral-nemo-base-2407": { - "max_tokens": 128000, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/mistral-nemo-instruct-2407": { - "max_tokens": 128000, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/mistral-small-24b-instruct-2501": { - "max_tokens": 32768, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/mixtral-8x22b": { - "max_tokens": 65536, + "input_cost_per_token": 1.2e-06, + "litellm_provider": "fireworks_ai", "max_input_tokens": 65536, "max_output_tokens": 65536, - "input_cost_per_token": 1.2e-06, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 1.2e-06 }, "fireworks_ai/accounts/fireworks/models/mixtral-8x22b-instruct": { - "max_tokens": 65536, + "input_cost_per_token": 1.2e-06, + "litellm_provider": "fireworks_ai", "max_input_tokens": 65536, "max_output_tokens": 65536, - "input_cost_per_token": 1.2e-06, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 1.2e-06 }, "fireworks_ai/accounts/fireworks/models/mixtral-8x7b": { - "max_tokens": 32768, + "input_cost_per_token": 5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 5e-07 }, "fireworks_ai/accounts/fireworks/models/mixtral-8x7b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 5e-07 }, "fireworks_ai/accounts/fireworks/models/mixtral-8x7b-instruct-hf": { - "max_tokens": 32768, + "input_cost_per_token": 5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 5e-07 }, "fireworks_ai/accounts/fireworks/models/mythomax-l2-13b": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/nemotron-nano-v2-12b-vl": { - "max_tokens": 4096, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/nous-capybara-7b-v1p9": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/nous-hermes-2-mixtral-8x7b-dpo": { - "max_tokens": 32768, + "input_cost_per_token": 5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 5e-07 }, "fireworks_ai/accounts/fireworks/models/nous-hermes-2-yi-34b": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/nous-hermes-llama2-13b": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/nous-hermes-llama2-70b": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/nous-hermes-llama2-7b": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/nvidia-nemotron-nano-12b-v2": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/nvidia-nemotron-nano-9b-v2": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/openchat-3p5-0106-7b": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/openhermes-2-mistral-7b": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/openhermes-2p5-mistral-7b": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/openorca-7b": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/phi-2-3b": { - "max_tokens": 2048, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 2048, "max_output_tokens": 2048, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 2048, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/phi-3-mini-128k-instruct": { - "max_tokens": 131072, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/phi-3-vision-128k-instruct": { - "max_tokens": 32064, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32064, "max_output_tokens": 32064, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32064, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/phind-code-llama-34b-python-v1": { - "max_tokens": 16384, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/phind-code-llama-34b-v1": { - "max_tokens": 16384, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/phind-code-llama-34b-v2": { - "max_tokens": 16384, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/playground-v2-1024px-aesthetic": { - "max_tokens": 4096, + "input_cost_per_token": 1.3e-10, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1.3e-10, - "output_cost_per_token": 1.3e-10, - "litellm_provider": "fireworks_ai", - "mode": "image_generation" + "max_tokens": 4096, + "mode": "image_generation", + "output_cost_per_token": 1.3e-10 }, "fireworks_ai/accounts/fireworks/models/playground-v2-5-1024px-aesthetic": { - "max_tokens": 4096, + "input_cost_per_token": 1.3e-10, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1.3e-10, - "output_cost_per_token": 1.3e-10, - "litellm_provider": "fireworks_ai", - "mode": "image_generation" + "max_tokens": 4096, + "mode": "image_generation", + "output_cost_per_token": 1.3e-10 }, "fireworks_ai/accounts/fireworks/models/pythia-12b": { - "max_tokens": 2048, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 2048, "max_output_tokens": 2048, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 2048, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen-qwq-32b-preview": { - "max_tokens": 32768, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen-v2p5-14b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen-v2p5-7b": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen1p5-72b-chat": { - "max_tokens": 32768, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2-7b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2-vl-2b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2-vl-72b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2-vl-7b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-0p5b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-14b": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-1p5b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-32b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-32b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-72b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-72b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-7b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-0p5b": { - "max_tokens": 32768, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-0p5b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-14b": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-14b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-1p5b": { - "max_tokens": 32768, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-1p5b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-32b": { - "max_tokens": 32768, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-32b-instruct-128k": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-32b-instruct-32k-rope": { - "max_tokens": 32768, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-32b-instruct-64k": { - "max_tokens": 65536, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 65536, "max_output_tokens": 65536, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-3b": { - "max_tokens": 32768, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-3b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-7b": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-7b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-math-72b-instruct": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-vl-32b-instruct": { - "max_tokens": 128000, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-vl-3b-instruct": { - "max_tokens": 128000, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-vl-72b-instruct": { - "max_tokens": 128000, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-vl-7b-instruct": { - "max_tokens": 128000, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-0p6b": { - "max_tokens": 40960, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 40960, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-14b": { - "max_tokens": 40960, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 40960, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-1p7b": { - "max_tokens": 131072, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-1p7b-fp8-draft": { - "max_tokens": 262144, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-1p7b-fp8-draft-131072": { - "max_tokens": 131072, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-1p7b-fp8-draft-40960": { - "max_tokens": 40960, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 40960, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-235b-a22b": { - "max_tokens": 131072, + "input_cost_per_token": 2.2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2.2e-07, - "output_cost_per_token": 8.8e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 8.8e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-235b-a22b-instruct-2507": { - "max_tokens": 262144, + "input_cost_per_token": 2.2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 2.2e-07, - "output_cost_per_token": 8.8e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 8.8e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-235b-a22b-thinking-2507": { - "max_tokens": 262144, + "input_cost_per_token": 2.2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 2.2e-07, - "output_cost_per_token": 8.8e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 8.8e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-30b-a3b": { - "max_tokens": 131072, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 6e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-30b-a3b-instruct-2507": { - "max_tokens": 262144, + "input_cost_per_token": 5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 5e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-30b-a3b-thinking-2507": { - "max_tokens": 262144, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-32b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 9e-07, "supports_reasoning": true }, "fireworks_ai/accounts/fireworks/models/qwen3-4b": { - "max_tokens": 40960, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 40960, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-4b-instruct-2507": { - "max_tokens": 262144, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-8b": { - "max_tokens": 40960, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", + "max_tokens": 40960, "mode": "chat", + "output_cost_per_token": 2e-07, "supports_reasoning": true }, "fireworks_ai/accounts/fireworks/models/qwen3-coder-30b-a3b-instruct": { - "max_tokens": 262144, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 6e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-coder-480b-instruct-bf16": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-embedding-0p6b": { - "max_tokens": 32768, + "input_cost_per_token": 0.0, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 0.0, - "output_cost_per_token": 0.0, - "litellm_provider": "fireworks_ai", - "mode": "embedding" + "max_tokens": 32768, + "mode": "embedding", + "output_cost_per_token": 0.0 }, "fireworks_ai/accounts/fireworks/models/qwen3-embedding-4b": { - "max_tokens": 40960, + "input_cost_per_token": 0.0, + "litellm_provider": "fireworks_ai", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 0.0, - "output_cost_per_token": 0.0, - "litellm_provider": "fireworks_ai", - "mode": "embedding" + "max_tokens": 40960, + "mode": "embedding", + "output_cost_per_token": 0.0 }, "fireworks_ai/accounts/fireworks/models/": { - "max_tokens": 40960, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 0.0, - "litellm_provider": "fireworks_ai", - "mode": "embedding" + "max_tokens": 40960, + "mode": "embedding", + "output_cost_per_token": 0.0 }, "fireworks_ai/accounts/fireworks/models/qwen3-next-80b-a3b-instruct": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-next-80b-a3b-thinking": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-reranker-0p6b": { - "max_tokens": 40960, + "input_cost_per_token": 0.0, + "litellm_provider": "fireworks_ai", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 0.0, - "output_cost_per_token": 0.0, - "litellm_provider": "fireworks_ai", - "mode": "rerank" + "max_tokens": 40960, + "mode": "rerank", + "output_cost_per_token": 0.0 }, "fireworks_ai/accounts/fireworks/models/qwen3-reranker-4b": { - "max_tokens": 40960, + "input_cost_per_token": 0.0, + "litellm_provider": "fireworks_ai", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 0.0, - "output_cost_per_token": 0.0, - "litellm_provider": "fireworks_ai", - "mode": "rerank" + "max_tokens": 40960, + "mode": "rerank", + "output_cost_per_token": 0.0 }, "fireworks_ai/accounts/fireworks/models/qwen3-reranker-8b": { - "max_tokens": 40960, + "input_cost_per_token": 0.0, + "litellm_provider": "fireworks_ai", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 0.0, - "output_cost_per_token": 0.0, - "litellm_provider": "fireworks_ai", - "mode": "rerank" + "max_tokens": 40960, + "mode": "rerank", + "output_cost_per_token": 0.0 }, "fireworks_ai/accounts/fireworks/models/qwen3-vl-235b-a22b-instruct": { - "max_tokens": 262144, + "input_cost_per_token": 2.2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 2.2e-07, - "output_cost_per_token": 8.8e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 8.8e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-vl-235b-a22b-thinking": { - "max_tokens": 262144, + "input_cost_per_token": 2.2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 2.2e-07, - "output_cost_per_token": 8.8e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 8.8e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-vl-30b-a3b-instruct": { - "max_tokens": 262144, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 6e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-vl-30b-a3b-thinking": { - "max_tokens": 262144, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 6e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-vl-32b-instruct": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-vl-8b-instruct": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3p7-plus": { "cache_read_input_token_cost": 8e-08, @@ -45991,148 +45995,148 @@ "supports_vision": true }, "fireworks_ai/accounts/fireworks/models/qwq-32b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/rolm-ocr": { - "max_tokens": 128000, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/snorkel-mistral-7b-pairrm-dpo": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/stable-diffusion-xl-1024-v1-0": { - "max_tokens": 4096, + "input_cost_per_token": 1.3e-10, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1.3e-10, - "output_cost_per_token": 1.3e-10, - "litellm_provider": "fireworks_ai", - "mode": "image_generation" + "max_tokens": 4096, + "mode": "image_generation", + "output_cost_per_token": 1.3e-10 }, "fireworks_ai/accounts/fireworks/models/stablecode-3b": { - "max_tokens": 4096, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/starcoder-16b": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/starcoder-7b": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/starcoder2-15b": { - "max_tokens": 16384, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/starcoder2-3b": { - "max_tokens": 16384, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/starcoder2-7b": { - "max_tokens": 16384, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/toppy-m-7b": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/yi-34b": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/yi-34b-200k-capybara": { - "max_tokens": 200000, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 200000, "max_output_tokens": 200000, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 200000, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/yi-34b-chat": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/yi-6b": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/zephyr-7b-beta": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/routers/glm-5p1-fast": { "cache_read_input_token_cost": 5.2e-07, @@ -46203,8 +46207,8 @@ "mode": "chat", "output_cost_per_token": 1.5e-06, "supports_function_calling": true, - "supports_vision": true, - "supports_reasoning": true + "supports_reasoning": true, + "supports_vision": true }, "scaleway/qwen/qwen3-235b-a22b-instruct-2507": { "input_cost_per_token": 7.5e-07, @@ -46279,10 +46283,10 @@ "max_tokens": 16384, "mode": "chat", "output_cost_per_token": 7.5e-06, - "supports_reasoning": true, "supports_function_calling": true, - "supports_vision": true, - "supports_tool_choice": true + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": true }, "scaleway/mistralai/mistral-small-3.2-24b-instruct-2506": { "input_cost_per_token": 1.5e-07, @@ -46303,8 +46307,8 @@ "max_tokens": 4096, "mode": "chat", "output_cost_per_token": 2e-07, - "supports_vision": true, - "supports_function_calling": true + "supports_function_calling": true, + "supports_vision": true }, "scaleway/BAAI/bge-multilingual-gemma2": { "input_cost_per_token": 1e-07, @@ -46323,1491 +46327,1491 @@ "supports_function_calling": true }, "novita/deepseek/deepseek-v3.2": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 1.345e-07, "input_cost_per_token": 2.69e-07, - "output_cost_per_token": 4e-07, + "input_cost_per_token_cache_hit": 1.345e-07, + "litellm_provider": "novita", "max_input_tokens": 163840, "max_output_tokens": 65536, "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 4e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "cache_read_input_token_cost": 1.345e-07, - "input_cost_per_token_cache_hit": 1.345e-07, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/minimax/minimax-m2.1": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 3e-08, "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, + "input_cost_per_token_cache_hit": 3e-08, + "litellm_provider": "novita", "max_input_tokens": 204800, "max_output_tokens": 131072, "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 1.2e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, "supports_response_schema": true, - "cache_read_input_token_cost": 3e-08, - "input_cost_per_token_cache_hit": 3e-08 + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/zai-org/glm-4.7": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 1.1e-07, "input_cost_per_token": 6e-07, - "output_cost_per_token": 2.2e-06, + "input_cost_per_token_cache_hit": 1.1e-07, + "litellm_provider": "novita", "max_input_tokens": 204800, "max_output_tokens": 131072, "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2.2e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "cache_read_input_token_cost": 1.1e-07, - "input_cost_per_token_cache_hit": 1.1e-07, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/xiaomimimo/mimo-v2-flash": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 2e-08, "input_cost_per_token": 1.1e-07, - "output_cost_per_token": 3.3e-07, + "input_cost_per_token_cache_hit": 2e-08, + "litellm_provider": "novita", "max_input_tokens": 262144, "max_output_tokens": 32000, "max_tokens": 32000, + "mode": "chat", + "output_cost_per_token": 3.3e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "cache_read_input_token_cost": 2e-08, - "input_cost_per_token_cache_hit": 2e-08, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/zai-org/autoglm-phone-9b-multilingual": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 3.5e-08, - "output_cost_per_token": 1.38e-07, + "litellm_provider": "novita", "max_input_tokens": 65536, "max_output_tokens": 65536, "max_tokens": 65536, - "supports_vision": true, - "supports_system_messages": true + "mode": "chat", + "output_cost_per_token": 1.38e-07, + "supports_system_messages": true, + "supports_vision": true }, "novita/moonshotai/kimi-k2-thinking": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 6e-07, - "output_cost_per_token": 2.5e-06, + "litellm_provider": "novita", "max_input_tokens": 262144, "max_output_tokens": 262144, "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 2.5e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/minimax/minimax-m2": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 3e-08, "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, + "input_cost_per_token_cache_hit": 3e-08, + "litellm_provider": "novita", "max_input_tokens": 204800, "max_output_tokens": 131072, "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 1.2e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, - "cache_read_input_token_cost": 3e-08, - "input_cost_per_token_cache_hit": 3e-08, "supports_reasoning": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/paddlepaddle/paddleocr-vl": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2e-08, - "output_cost_per_token": 2e-08, + "litellm_provider": "novita", "max_input_tokens": 16384, "max_output_tokens": 16384, "max_tokens": 16384, - "supports_vision": true, - "supports_system_messages": true + "mode": "chat", + "output_cost_per_token": 2e-08, + "supports_system_messages": true, + "supports_vision": true }, "novita/deepseek/deepseek-v3.2-exp": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2.7e-07, - "output_cost_per_token": 4.1e-07, + "litellm_provider": "novita", "max_input_tokens": 163840, "max_output_tokens": 65536, "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 4.1e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, - "supports_response_schema": true, - "supports_reasoning": true - }, - "novita/qwen/qwen3-vl-235b-a22b-thinking": { - "litellm_provider": "novita", - "mode": "chat", - "input_cost_per_token": 9.8e-07, - "output_cost_per_token": 3.95e-06, - "max_input_tokens": 131072, - "max_output_tokens": 32768, - "max_tokens": 32768, - "supports_vision": true, - "supports_system_messages": true, "supports_reasoning": true, - "supports_function_calling": true, + "supports_response_schema": true, + "supports_system_messages": true, "supports_tool_choice": true }, - "novita/zai-org/glm-4.6v": { + "novita/qwen/qwen3-vl-235b-a22b-thinking": { + "input_cost_per_token": 9.8e-07, "litellm_provider": "novita", - "mode": "chat", - "input_cost_per_token": 3e-07, - "output_cost_per_token": 9e-07, "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 3.95e-06, + "supports_function_calling": true, + "supports_reasoning": true, + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "novita/zai-org/glm-4.6v": { + "cache_read_input_token_cost": 5.5e-08, + "input_cost_per_token": 3e-07, + "input_cost_per_token_cache_hit": 5.5e-08, + "litellm_provider": "novita", + "max_input_tokens": 131072, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 9e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_vision": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "cache_read_input_token_cost": 5.5e-08, - "input_cost_per_token_cache_hit": 5.5e-08, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true }, "novita/zai-org/glm-4.6": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 1.1e-07, "input_cost_per_token": 5.5e-07, - "output_cost_per_token": 2.2e-06, + "input_cost_per_token_cache_hit": 1.1e-07, + "litellm_provider": "novita", "max_input_tokens": 204800, "max_output_tokens": 131072, "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2.2e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "cache_read_input_token_cost": 1.1e-07, - "input_cost_per_token_cache_hit": 1.1e-07, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/kwaipilot/kat-coder-pro": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 6e-08, "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, + "input_cost_per_token_cache_hit": 6e-08, + "litellm_provider": "novita", "max_input_tokens": 256000, "max_output_tokens": 128000, "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1.2e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, "supports_response_schema": true, - "cache_read_input_token_cost": 6e-08, - "input_cost_per_token_cache_hit": 6e-08 + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/qwen/qwen3-next-80b-a3b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 1.5e-06, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1.5e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/qwen/qwen3-next-80b-a3b-thinking": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 1.5e-06, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1.5e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/deepseek/deepseek-ocr": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 3e-08, - "output_cost_per_token": 3e-08, + "litellm_provider": "novita", "max_input_tokens": 8192, "max_output_tokens": 8192, "max_tokens": 8192, - "supports_tool_choice": true, - "supports_vision": true, + "mode": "chat", + "output_cost_per_token": 3e-08, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true, + "supports_vision": true }, "novita/deepseek/deepseek-v3.1-terminus": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 1.35e-07, "input_cost_per_token": 2.7e-07, - "output_cost_per_token": 1e-06, + "input_cost_per_token_cache_hit": 1.35e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "cache_read_input_token_cost": 1.35e-07, - "input_cost_per_token_cache_hit": 1.35e-07, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/qwen/qwen3-vl-235b-a22b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.5e-06, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1.5e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true, + "supports_vision": true }, "novita/qwen/qwen3-max": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2.11e-06, - "output_cost_per_token": 8.45e-06, + "litellm_provider": "novita", "max_input_tokens": 262144, "max_output_tokens": 65536, "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 8.45e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/skywork/r1v4-lite": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2e-07, - "output_cost_per_token": 6e-07, + "litellm_provider": "novita", "max_input_tokens": 262144, "max_output_tokens": 65536, "max_tokens": 65536, - "supports_tool_choice": true, - "supports_vision": true, + "mode": "chat", + "output_cost_per_token": 6e-07, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true, + "supports_vision": true }, "novita/deepseek/deepseek-v3.1": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 1.35e-07, "input_cost_per_token": 2.7e-07, - "output_cost_per_token": 1e-06, + "input_cost_per_token_cache_hit": 1.35e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "cache_read_input_token_cost": 1.35e-07, - "input_cost_per_token_cache_hit": 1.35e-07, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/moonshotai/kimi-k2-0905": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 6e-07, - "output_cost_per_token": 2.5e-06, + "litellm_provider": "novita", "max_input_tokens": 262144, "max_output_tokens": 262144, "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 2.5e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/qwen/qwen3-coder-480b-a35b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 3.8e-07, - "output_cost_per_token": 1.55e-06, + "litellm_provider": "novita", "max_input_tokens": 262144, "max_output_tokens": 65536, "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 1.55e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/qwen/qwen3-coder-30b-a3b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 7e-08, - "output_cost_per_token": 2.7e-07, + "litellm_provider": "novita", "max_input_tokens": 160000, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2.7e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/openai/gpt-oss-120b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 5e-08, - "output_cost_per_token": 2.5e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2.5e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_vision": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true }, "novita/moonshotai/kimi-k2-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 5.7e-07, - "output_cost_per_token": 2.3e-06, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 131072, "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2.3e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/deepseek/deepseek-v3-0324": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 1.35e-07, "input_cost_per_token": 2.7e-07, - "output_cost_per_token": 1.12e-06, + "input_cost_per_token_cache_hit": 1.35e-07, + "litellm_provider": "novita", "max_input_tokens": 163840, "max_output_tokens": 163840, "max_tokens": 163840, + "mode": "chat", + "output_cost_per_token": 1.12e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, "supports_response_schema": true, - "cache_read_input_token_cost": 1.35e-07, - "input_cost_per_token_cache_hit": 1.35e-07 + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/zai-org/glm-4.5": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 1.1e-07, "input_cost_per_token": 6e-07, - "output_cost_per_token": 2.2e-06, + "input_cost_per_token_cache_hit": 1.1e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 98304, "max_tokens": 98304, + "mode": "chat", + "output_cost_per_token": 2.2e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_reasoning": true, "supports_system_messages": true, - "cache_read_input_token_cost": 1.1e-07, - "input_cost_per_token_cache_hit": 1.1e-07, - "supports_reasoning": true + "supports_tool_choice": true }, "novita/qwen/qwen3-235b-a22b-thinking-2507": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 3e-07, - "output_cost_per_token": 3e-06, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 3e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_reasoning": true, "supports_system_messages": true, - "supports_reasoning": true + "supports_tool_choice": true }, "novita/meta-llama/llama-3.1-8b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2e-08, - "output_cost_per_token": 5e-08, + "litellm_provider": "novita", "max_input_tokens": 16384, "max_output_tokens": 16384, "max_tokens": 16384, - "supports_system_messages": true, - "supports_response_schema": true + "mode": "chat", + "output_cost_per_token": 5e-08, + "supports_response_schema": true, + "supports_system_messages": true }, "novita/google/gemma-3-12b-it": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 5e-08, - "output_cost_per_token": 1e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 8192, "max_tokens": 8192, - "supports_tool_choice": true, - "supports_vision": true, + "mode": "chat", + "output_cost_per_token": 1e-07, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true, + "supports_vision": true }, "novita/zai-org/glm-4.5v": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 1.1e-07, "input_cost_per_token": 6e-07, - "output_cost_per_token": 1.8e-06, + "input_cost_per_token_cache_hit": 1.1e-07, + "litellm_provider": "novita", "max_input_tokens": 65536, "max_output_tokens": 16384, "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 1.8e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_vision": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "cache_read_input_token_cost": 1.1e-07, - "input_cost_per_token_cache_hit": 1.1e-07, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true }, "novita/openai/gpt-oss-20b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 4e-08, - "output_cost_per_token": 1.5e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, - "supports_tool_choice": true, - "supports_vision": true, - "supports_system_messages": true, + "mode": "chat", + "output_cost_per_token": 1.5e-07, + "supports_reasoning": true, "supports_response_schema": true, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true }, "novita/qwen/qwen3-235b-a22b-instruct-2507": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 9e-08, - "output_cost_per_token": 5.8e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 16384, "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 5.8e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/deepseek/deepseek-r1-distill-qwen-14b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 1.5e-07, + "litellm_provider": "novita", "max_input_tokens": 32768, "max_output_tokens": 16384, "max_tokens": 16384, - "supports_tool_choice": true, - "supports_system_messages": true, + "mode": "chat", + "output_cost_per_token": 1.5e-07, + "supports_reasoning": true, "supports_response_schema": true, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/meta-llama/llama-3.3-70b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1.35e-07, - "output_cost_per_token": 4e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 120000, "max_tokens": 120000, + "mode": "chat", + "output_cost_per_token": 4e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/qwen/qwen-2.5-72b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 3.8e-07, - "output_cost_per_token": 4e-07, + "litellm_provider": "novita", "max_input_tokens": 32000, "max_output_tokens": 8192, "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 4e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/mistralai/mistral-nemo": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 4e-08, - "output_cost_per_token": 1.7e-07, + "litellm_provider": "novita", "max_input_tokens": 60288, "max_output_tokens": 16000, "max_tokens": 16000, - "supports_tool_choice": true, + "mode": "chat", + "output_cost_per_token": 1.7e-07, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/minimaxai/minimax-m1-80k": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 5.5e-07, - "output_cost_per_token": 2.2e-06, + "litellm_provider": "novita", "max_input_tokens": 1000000, "max_output_tokens": 40000, "max_tokens": 40000, + "mode": "chat", + "output_cost_per_token": 2.2e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, "supports_reasoning": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/deepseek/deepseek-r1-0528": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 3.5e-07, "input_cost_per_token": 7e-07, - "output_cost_per_token": 2.5e-06, + "input_cost_per_token_cache_hit": 3.5e-07, + "litellm_provider": "novita", "max_input_tokens": 163840, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2.5e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "cache_read_input_token_cost": 3.5e-07, - "input_cost_per_token_cache_hit": 3.5e-07, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/deepseek/deepseek-r1-distill-qwen-32b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 3e-07, - "output_cost_per_token": 3e-07, + "litellm_provider": "novita", "max_input_tokens": 64000, "max_output_tokens": 32000, "max_tokens": 32000, - "supports_tool_choice": true, - "supports_system_messages": true, + "mode": "chat", + "output_cost_per_token": 3e-07, + "supports_reasoning": true, "supports_response_schema": true, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/meta-llama/llama-3-8b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 4e-08, - "output_cost_per_token": 4e-08, + "litellm_provider": "novita", "max_input_tokens": 8192, "max_output_tokens": 8192, "max_tokens": 8192, - "supports_system_messages": true, - "supports_response_schema": true + "mode": "chat", + "output_cost_per_token": 4e-08, + "supports_response_schema": true, + "supports_system_messages": true }, "novita/microsoft/wizardlm-2-8x22b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 6.2e-07, - "output_cost_per_token": 6.2e-07, + "litellm_provider": "novita", "max_input_tokens": 65535, "max_output_tokens": 8000, "max_tokens": 8000, - "supports_system_messages": true, - "supports_response_schema": true + "mode": "chat", + "output_cost_per_token": 6.2e-07, + "supports_response_schema": true, + "supports_system_messages": true }, "novita/deepseek/deepseek-r1-0528-qwen3-8b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 6e-08, - "output_cost_per_token": 9e-08, + "litellm_provider": "novita", "max_input_tokens": 128000, "max_output_tokens": 32000, "max_tokens": 32000, - "supports_system_messages": true, - "supports_reasoning": true + "mode": "chat", + "output_cost_per_token": 9e-08, + "supports_reasoning": true, + "supports_system_messages": true }, "novita/deepseek/deepseek-r1-distill-llama-70b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 8e-07, - "output_cost_per_token": 8e-07, + "litellm_provider": "novita", "max_input_tokens": 8192, "max_output_tokens": 8192, "max_tokens": 8192, - "supports_tool_choice": true, - "supports_system_messages": true, + "mode": "chat", + "output_cost_per_token": 8e-07, + "supports_reasoning": true, "supports_response_schema": true, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/meta-llama/llama-3-70b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 5.1e-07, - "output_cost_per_token": 7.4e-07, + "litellm_provider": "novita", "max_input_tokens": 8192, "max_output_tokens": 8000, "max_tokens": 8000, - "supports_tool_choice": true, + "mode": "chat", + "output_cost_per_token": 7.4e-07, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/qwen/qwen3-235b-a22b-fp8": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2e-07, - "output_cost_per_token": 8e-07, + "litellm_provider": "novita", "max_input_tokens": 40960, "max_output_tokens": 20000, "max_tokens": 20000, - "supports_system_messages": true, + "mode": "chat", + "output_cost_per_token": 8e-07, "supports_reasoning": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_system_messages": true }, "novita/meta-llama/llama-4-maverick-17b-128e-instruct-fp8": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2.7e-07, - "output_cost_per_token": 8.5e-07, + "litellm_provider": "novita", "max_input_tokens": 1048576, "max_output_tokens": 8192, "max_tokens": 8192, - "supports_vision": true, + "mode": "chat", + "output_cost_per_token": 8.5e-07, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_vision": true }, "novita/meta-llama/llama-4-scout-17b-16e-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1.8e-07, - "output_cost_per_token": 5.9e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 131072, "max_tokens": 131072, - "supports_vision": true, - "supports_system_messages": true + "mode": "chat", + "output_cost_per_token": 5.9e-07, + "supports_system_messages": true, + "supports_vision": true }, "novita/nousresearch/hermes-2-pro-llama-3-8b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1.4e-07, - "output_cost_per_token": 1.4e-07, + "litellm_provider": "novita", "max_input_tokens": 8192, "max_output_tokens": 8192, "max_tokens": 8192, - "supports_tool_choice": true, + "mode": "chat", + "output_cost_per_token": 1.4e-07, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/qwen/qwen2.5-vl-72b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 8e-07, - "output_cost_per_token": 8e-07, + "litellm_provider": "novita", "max_input_tokens": 32768, "max_output_tokens": 32768, "max_tokens": 32768, - "supports_vision": true, - "supports_system_messages": true + "mode": "chat", + "output_cost_per_token": 8e-07, + "supports_system_messages": true, + "supports_vision": true }, "novita/sao10k/l3-70b-euryale-v2.1": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1.48e-06, - "output_cost_per_token": 1.48e-06, + "litellm_provider": "novita", "max_input_tokens": 8192, "max_output_tokens": 8192, "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 1.48e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/baidu/ernie-4.5-21B-a3b-thinking": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 7e-08, - "output_cost_per_token": 2.8e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 65536, "max_tokens": 65536, - "supports_system_messages": true, - "supports_reasoning": true + "mode": "chat", + "output_cost_per_token": 2.8e-07, + "supports_reasoning": true, + "supports_system_messages": true }, "novita/sao10k/l3-8b-lunaris": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 5e-08, - "output_cost_per_token": 5e-08, + "litellm_provider": "novita", "max_input_tokens": 8192, "max_output_tokens": 8192, "max_tokens": 8192, - "supports_tool_choice": true, + "mode": "chat", + "output_cost_per_token": 5e-08, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/baichuan/baichuan-m2-32b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 7e-08, - "output_cost_per_token": 7e-08, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 131072, "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 7e-08, "supports_system_messages": true }, "novita/baidu/ernie-4.5-vl-424b-a47b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 4.2e-07, - "output_cost_per_token": 1.25e-06, + "litellm_provider": "novita", "max_input_tokens": 123000, "max_output_tokens": 16000, "max_tokens": 16000, - "supports_vision": true, + "mode": "chat", + "output_cost_per_token": 1.25e-06, + "supports_reasoning": true, "supports_system_messages": true, - "supports_reasoning": true + "supports_vision": true }, "novita/baidu/ernie-4.5-300b-a47b-paddle": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2.8e-07, - "output_cost_per_token": 1.1e-06, + "litellm_provider": "novita", "max_input_tokens": 123000, "max_output_tokens": 12000, "max_tokens": 12000, - "supports_tool_choice": true, + "mode": "chat", + "output_cost_per_token": 1.1e-06, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/deepseek/deepseek-prover-v2-671b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 7e-07, - "output_cost_per_token": 2.5e-06, + "litellm_provider": "novita", "max_input_tokens": 160000, "max_output_tokens": 160000, "max_tokens": 160000, + "mode": "chat", + "output_cost_per_token": 2.5e-06, "supports_system_messages": true }, "novita/qwen/qwen3-32b-fp8": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1e-07, - "output_cost_per_token": 4.5e-07, + "litellm_provider": "novita", "max_input_tokens": 40960, "max_output_tokens": 20000, "max_tokens": 20000, - "supports_system_messages": true, - "supports_reasoning": true + "mode": "chat", + "output_cost_per_token": 4.5e-07, + "supports_reasoning": true, + "supports_system_messages": true }, "novita/qwen/qwen3-30b-a3b-fp8": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 9e-08, - "output_cost_per_token": 4.5e-07, + "litellm_provider": "novita", "max_input_tokens": 40960, "max_output_tokens": 20000, "max_tokens": 20000, - "supports_system_messages": true, - "supports_reasoning": true, + "mode": "chat", + "output_cost_per_token": 4.5e-07, "supports_function_calling": true, + "supports_reasoning": true, + "supports_system_messages": true, "supports_tool_choice": true }, "novita/google/gemma-3-27b-it": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1.19e-07, - "output_cost_per_token": 2e-07, + "litellm_provider": "novita", "max_input_tokens": 98304, "max_output_tokens": 16384, "max_tokens": 16384, - "supports_vision": true, - "supports_system_messages": true + "mode": "chat", + "output_cost_per_token": 2e-07, + "supports_system_messages": true, + "supports_vision": true }, "novita/deepseek/deepseek-v3-turbo": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 4e-07, - "output_cost_per_token": 1.3e-06, + "litellm_provider": "novita", "max_input_tokens": 64000, "max_output_tokens": 16000, "max_tokens": 16000, + "mode": "chat", + "output_cost_per_token": 1.3e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/deepseek/deepseek-r1-turbo": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 7e-07, - "output_cost_per_token": 2.5e-06, + "litellm_provider": "novita", "max_input_tokens": 64000, "max_output_tokens": 16000, "max_tokens": 16000, + "mode": "chat", + "output_cost_per_token": 2.5e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, "supports_reasoning": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/Sao10K/L3-8B-Stheno-v3.2": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 5e-08, - "output_cost_per_token": 5e-08, + "litellm_provider": "novita", "max_input_tokens": 8192, "max_output_tokens": 32000, "max_tokens": 32000, + "mode": "chat", + "output_cost_per_token": 5e-08, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/gryphe/mythomax-l2-13b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 9e-08, - "output_cost_per_token": 9e-08, + "litellm_provider": "novita", "max_input_tokens": 4096, "max_output_tokens": 3200, "max_tokens": 3200, + "mode": "chat", + "output_cost_per_token": 9e-08, "supports_system_messages": true }, "novita/baidu/ernie-4.5-vl-28b-a3b-thinking": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 3.9e-07, - "output_cost_per_token": 3.9e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 65536, "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 3.9e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_vision": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true }, "novita/qwen/qwen3-vl-8b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 8e-08, - "output_cost_per_token": 5e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 5e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true, + "supports_vision": true }, "novita/zai-org/glm-4.5-air": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1.3e-07, - "output_cost_per_token": 8.5e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 98304, "max_tokens": 98304, + "mode": "chat", + "output_cost_per_token": 8.5e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_reasoning": true, "supports_system_messages": true, - "supports_reasoning": true + "supports_tool_choice": true }, "novita/qwen/qwen3-vl-30b-a3b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2e-07, - "output_cost_per_token": 7e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 7e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true, + "supports_vision": true }, "novita/qwen/qwen3-vl-30b-a3b-thinking": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2e-07, - "output_cost_per_token": 1e-06, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true, + "supports_vision": true }, "novita/qwen/qwen3-omni-30b-a3b-thinking": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2.5e-07, - "output_cost_per_token": 9.7e-07, + "litellm_provider": "novita", "max_input_tokens": 65536, "max_output_tokens": 16384, "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 9.7e-07, + "supports_audio_input": true, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_vision": true, - "supports_system_messages": true, - "supports_response_schema": true, "supports_reasoning": true, - "supports_audio_input": true + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true }, "novita/qwen/qwen3-omni-30b-a3b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2.5e-07, - "output_cost_per_token": 9.7e-07, + "litellm_provider": "novita", "max_input_tokens": 65536, "max_output_tokens": 16384, "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 9.7e-07, + "supports_audio_input": true, + "supports_audio_output": true, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_vision": true, - "supports_system_messages": true, "supports_response_schema": true, - "supports_audio_input": true, - "supports_audio_output": true + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true }, "novita/qwen/qwen-mt-plus": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2.5e-07, - "output_cost_per_token": 7.5e-07, + "litellm_provider": "novita", "max_input_tokens": 16384, "max_output_tokens": 8192, "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 7.5e-07, "supports_system_messages": true }, "novita/baidu/ernie-4.5-vl-28b-a3b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1.4e-07, - "output_cost_per_token": 5.6e-07, + "litellm_provider": "novita", "max_input_tokens": 30000, "max_output_tokens": 8000, "max_tokens": 8000, + "mode": "chat", + "output_cost_per_token": 5.6e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_vision": true, + "supports_reasoning": true, "supports_system_messages": true, - "supports_reasoning": true + "supports_tool_choice": true, + "supports_vision": true }, "novita/baidu/ernie-4.5-21B-a3b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 7e-08, - "output_cost_per_token": 2.8e-07, + "litellm_provider": "novita", "max_input_tokens": 120000, "max_output_tokens": 8000, "max_tokens": 8000, + "mode": "chat", + "output_cost_per_token": 2.8e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/qwen/qwen3-8b-fp8": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 3.5e-08, - "output_cost_per_token": 1.38e-07, + "litellm_provider": "novita", "max_input_tokens": 128000, "max_output_tokens": 20000, "max_tokens": 20000, - "supports_system_messages": true, - "supports_reasoning": true + "mode": "chat", + "output_cost_per_token": 1.38e-07, + "supports_reasoning": true, + "supports_system_messages": true }, "novita/qwen/qwen3-4b-fp8": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 3e-08, - "output_cost_per_token": 3e-08, + "litellm_provider": "novita", "max_input_tokens": 128000, "max_output_tokens": 20000, "max_tokens": 20000, - "supports_system_messages": true, - "supports_reasoning": true, + "mode": "chat", + "output_cost_per_token": 3e-08, "supports_function_calling": true, + "supports_reasoning": true, + "supports_system_messages": true, "supports_tool_choice": true }, "novita/qwen/qwen2.5-7b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 7e-08, - "output_cost_per_token": 7e-08, + "litellm_provider": "novita", "max_input_tokens": 32000, "max_output_tokens": 32000, "max_tokens": 32000, + "mode": "chat", + "output_cost_per_token": 7e-08, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/meta-llama/llama-3.2-3b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 3e-08, - "output_cost_per_token": 5e-08, + "litellm_provider": "novita", "max_input_tokens": 32768, "max_output_tokens": 32000, "max_tokens": 32000, + "mode": "chat", + "output_cost_per_token": 5e-08, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/sao10k/l31-70b-euryale-v2.2": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1.48e-06, - "output_cost_per_token": 1.48e-06, + "litellm_provider": "novita", "max_input_tokens": 8192, "max_output_tokens": 8192, "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 1.48e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/qwen/qwen3-embedding-0.6b": { - "litellm_provider": "novita", - "mode": "embedding", "input_cost_per_token": 7e-08, - "output_cost_per_token": 0, + "litellm_provider": "novita", "max_input_tokens": 32768, "max_output_tokens": 32768, - "max_tokens": 32768 + "max_tokens": 32768, + "mode": "embedding", + "output_cost_per_token": 0 }, "novita/qwen/qwen3-embedding-8b": { - "litellm_provider": "novita", - "mode": "embedding", "input_cost_per_token": 7e-08, - "output_cost_per_token": 0, + "litellm_provider": "novita", "max_input_tokens": 32768, "max_output_tokens": 4096, - "max_tokens": 4096 + "max_tokens": 4096, + "mode": "embedding", + "output_cost_per_token": 0 }, "novita/baai/bge-m3": { - "litellm_provider": "novita", - "mode": "embedding", "input_cost_per_token": 1e-08, - "output_cost_per_token": 1e-08, + "litellm_provider": "novita", "max_input_tokens": 8192, "max_output_tokens": 96000, - "max_tokens": 96000 + "max_tokens": 96000, + "mode": "embedding", + "output_cost_per_token": 1e-08 }, "novita/qwen/qwen3-reranker-8b": { - "litellm_provider": "novita", - "mode": "rerank", "input_cost_per_token": 5e-08, - "output_cost_per_token": 5e-08, + "litellm_provider": "novita", "max_input_tokens": 32768, "max_output_tokens": 4096, - "max_tokens": 4096 + "max_tokens": 4096, + "mode": "rerank", + "output_cost_per_token": 5e-08 }, "novita/baai/bge-reranker-v2-m3": { - "litellm_provider": "novita", - "mode": "rerank", "input_cost_per_token": 1e-08, - "output_cost_per_token": 1e-08, + "litellm_provider": "novita", "max_input_tokens": 8000, "max_output_tokens": 8000, - "max_tokens": 8000 + "max_tokens": 8000, + "mode": "rerank", + "output_cost_per_token": 1e-08 }, "llamagate/llama-3.1-8b": { - "max_tokens": 8192, + "input_cost_per_token": 3e-08, + "litellm_provider": "llamagate", "max_input_tokens": 131072, "max_output_tokens": 8192, - "input_cost_per_token": 3e-08, - "output_cost_per_token": 5e-08, - "litellm_provider": "llamagate", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 5e-08, "supports_function_calling": true, "supports_response_schema": true }, "llamagate/llama-3.2-3b": { - "max_tokens": 8192, + "input_cost_per_token": 4e-08, + "litellm_provider": "llamagate", "max_input_tokens": 131072, "max_output_tokens": 8192, - "input_cost_per_token": 4e-08, - "output_cost_per_token": 8e-08, - "litellm_provider": "llamagate", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 8e-08, "supports_function_calling": true, "supports_response_schema": true }, "llamagate/mistral-7b-v0.3": { - "max_tokens": 8192, + "input_cost_per_token": 1e-07, + "litellm_provider": "llamagate", "max_input_tokens": 32768, "max_output_tokens": 8192, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1.5e-07, - "litellm_provider": "llamagate", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.5e-07, "supports_function_calling": true, "supports_response_schema": true }, "llamagate/qwen3-8b": { - "max_tokens": 8192, + "input_cost_per_token": 4e-08, + "litellm_provider": "llamagate", "max_input_tokens": 32768, "max_output_tokens": 8192, - "input_cost_per_token": 4e-08, - "output_cost_per_token": 1.4e-07, - "litellm_provider": "llamagate", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.4e-07, "supports_function_calling": true, "supports_response_schema": true }, "llamagate/dolphin3-8b": { - "max_tokens": 8192, + "input_cost_per_token": 8e-08, + "litellm_provider": "llamagate", "max_input_tokens": 128000, "max_output_tokens": 8192, - "input_cost_per_token": 8e-08, - "output_cost_per_token": 1.5e-07, - "litellm_provider": "llamagate", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.5e-07, "supports_function_calling": true, "supports_response_schema": true }, "llamagate/deepseek-r1-8b": { - "max_tokens": 16384, + "input_cost_per_token": 1e-07, + "litellm_provider": "llamagate", "max_input_tokens": 65536, "max_output_tokens": 16384, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "llamagate", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 2e-07, "supports_function_calling": true, - "supports_response_schema": true, - "supports_reasoning": true + "supports_reasoning": true, + "supports_response_schema": true }, "llamagate/deepseek-r1-7b-qwen": { - "max_tokens": 16384, + "input_cost_per_token": 8e-08, + "litellm_provider": "llamagate", "max_input_tokens": 131072, "max_output_tokens": 16384, - "input_cost_per_token": 8e-08, - "output_cost_per_token": 1.5e-07, - "litellm_provider": "llamagate", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 1.5e-07, "supports_function_calling": true, - "supports_response_schema": true, - "supports_reasoning": true + "supports_reasoning": true, + "supports_response_schema": true }, "llamagate/openthinker-7b": { - "max_tokens": 8192, + "input_cost_per_token": 8e-08, + "litellm_provider": "llamagate", "max_input_tokens": 32768, "max_output_tokens": 8192, - "input_cost_per_token": 8e-08, - "output_cost_per_token": 1.5e-07, - "litellm_provider": "llamagate", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.5e-07, "supports_function_calling": true, - "supports_response_schema": true, - "supports_reasoning": true + "supports_reasoning": true, + "supports_response_schema": true }, "llamagate/qwen2.5-coder-7b": { - "max_tokens": 8192, + "input_cost_per_token": 6e-08, + "litellm_provider": "llamagate", "max_input_tokens": 32768, "max_output_tokens": 8192, - "input_cost_per_token": 6e-08, - "output_cost_per_token": 1.2e-07, - "litellm_provider": "llamagate", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.2e-07, "supports_function_calling": true, "supports_response_schema": true }, "llamagate/deepseek-coder-6.7b": { - "max_tokens": 4096, + "input_cost_per_token": 6e-08, + "litellm_provider": "llamagate", "max_input_tokens": 16384, "max_output_tokens": 4096, - "input_cost_per_token": 6e-08, - "output_cost_per_token": 1.2e-07, - "litellm_provider": "llamagate", + "max_tokens": 4096, "mode": "chat", + "output_cost_per_token": 1.2e-07, "supports_function_calling": true, "supports_response_schema": true }, "llamagate/codellama-7b": { - "max_tokens": 4096, + "input_cost_per_token": 6e-08, + "litellm_provider": "llamagate", "max_input_tokens": 16384, "max_output_tokens": 4096, - "input_cost_per_token": 6e-08, - "output_cost_per_token": 1.2e-07, - "litellm_provider": "llamagate", + "max_tokens": 4096, "mode": "chat", + "output_cost_per_token": 1.2e-07, "supports_function_calling": true, "supports_response_schema": true }, "llamagate/qwen3-vl-8b": { - "max_tokens": 8192, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "llamagate", "max_input_tokens": 32768, "max_output_tokens": 8192, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 5.5e-07, - "litellm_provider": "llamagate", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 5.5e-07, "supports_function_calling": true, "supports_response_schema": true, "supports_vision": true }, "llamagate/llava-7b": { - "max_tokens": 2048, + "input_cost_per_token": 1e-07, + "litellm_provider": "llamagate", "max_input_tokens": 4096, "max_output_tokens": 2048, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "llamagate", + "max_tokens": 2048, "mode": "chat", + "output_cost_per_token": 2e-07, "supports_response_schema": true, "supports_vision": true }, "llamagate/gemma3-4b": { - "max_tokens": 8192, + "input_cost_per_token": 3e-08, + "litellm_provider": "llamagate", "max_input_tokens": 128000, "max_output_tokens": 8192, - "input_cost_per_token": 3e-08, - "output_cost_per_token": 8e-08, - "litellm_provider": "llamagate", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 8e-08, "supports_function_calling": true, "supports_response_schema": true, "supports_vision": true }, "llamagate/nomic-embed-text": { - "max_tokens": 8192, - "max_input_tokens": 8192, "input_cost_per_token": 2e-08, - "output_cost_per_token": 0, "litellm_provider": "llamagate", - "mode": "embedding" + "max_input_tokens": 8192, + "max_tokens": 8192, + "mode": "embedding", + "output_cost_per_token": 0 }, "llamagate/qwen3-embedding-8b": { - "max_tokens": 40960, - "max_input_tokens": 40960, "input_cost_per_token": 2e-08, - "output_cost_per_token": 0, "litellm_provider": "llamagate", - "mode": "embedding" + "max_input_tokens": 40960, + "max_tokens": 40960, + "mode": "embedding", + "output_cost_per_token": 0 }, "libertai/hermes-3-8b-tee": { - "max_tokens": 16000, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "libertai", "max_input_tokens": 16000, "max_output_tokens": 16000, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "libertai", + "max_tokens": 16000, "mode": "chat", + "output_cost_per_token": 6e-07, + "source": "https://docs.libertai.io/apis/text/", "supports_function_calling": true, - "supports_tool_choice": true, "supports_system_messages": true, - "supports_vision": false, - "source": "https://docs.libertai.io/apis/text/" + "supports_tool_choice": true, + "supports_vision": false }, "libertai/gemma-4-31b-it": { - "max_tokens": 262144, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "libertai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "libertai", + "max_tokens": 262144, "mode": "chat", + "output_cost_per_token": 4e-07, + "source": "https://docs.libertai.io/apis/text/", "supports_function_calling": true, - "supports_tool_choice": true, "supports_system_messages": true, - "supports_vision": true, - "source": "https://docs.libertai.io/apis/text/" + "supports_tool_choice": true, + "supports_vision": true }, "libertai/gemma-4-31b-it-thinking": { - "max_tokens": 262144, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "libertai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "libertai", + "max_tokens": 262144, "mode": "chat", + "output_cost_per_token": 4e-07, + "source": "https://docs.libertai.io/apis/text/", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, - "supports_vision": true, "supports_reasoning": true, - "source": "https://docs.libertai.io/apis/text/" + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true }, "libertai/qwen3.6-27b": { - "max_tokens": 262144, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "libertai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "libertai", + "max_tokens": 262144, "mode": "chat", + "output_cost_per_token": 5e-07, + "source": "https://docs.libertai.io/apis/text/", "supports_function_calling": true, - "supports_tool_choice": true, "supports_system_messages": true, - "supports_vision": true, - "source": "https://docs.libertai.io/apis/text/" + "supports_tool_choice": true, + "supports_vision": true }, "libertai/qwen3.6-27b-thinking": { - "max_tokens": 262144, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "libertai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "libertai", + "max_tokens": 262144, "mode": "chat", + "output_cost_per_token": 5e-07, + "source": "https://docs.libertai.io/apis/text/", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, - "supports_vision": true, "supports_reasoning": true, - "source": "https://docs.libertai.io/apis/text/" + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true }, "libertai/qwen3.6-35b-a3b": { - "max_tokens": 262144, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "libertai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "libertai", + "max_tokens": 262144, "mode": "chat", + "output_cost_per_token": 5e-07, + "source": "https://docs.libertai.io/apis/text/", "supports_function_calling": true, - "supports_tool_choice": true, "supports_system_messages": true, - "supports_vision": true, - "source": "https://docs.libertai.io/apis/text/" + "supports_tool_choice": true, + "supports_vision": true }, "libertai/qwen3.6-35b-a3b-thinking": { - "max_tokens": 262144, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "libertai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "libertai", + "max_tokens": 262144, "mode": "chat", + "output_cost_per_token": 5e-07, + "source": "https://docs.libertai.io/apis/text/", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, - "supports_vision": true, "supports_reasoning": true, - "source": "https://docs.libertai.io/apis/text/" + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true }, "libertai/qwen3.5-122b-a10b": { - "max_tokens": 262144, + "input_cost_per_token": 2.5e-07, + "litellm_provider": "libertai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 2.5e-07, - "output_cost_per_token": 1.75e-06, - "litellm_provider": "libertai", + "max_tokens": 262144, "mode": "chat", + "output_cost_per_token": 1.75e-06, + "source": "https://docs.libertai.io/apis/text/", "supports_function_calling": true, - "supports_tool_choice": true, "supports_system_messages": true, - "supports_vision": true, - "source": "https://docs.libertai.io/apis/text/" + "supports_tool_choice": true, + "supports_vision": true }, "libertai/qwen3.5-122b-a10b-thinking": { - "max_tokens": 262144, + "input_cost_per_token": 2.5e-07, + "litellm_provider": "libertai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 2.5e-07, - "output_cost_per_token": 1.75e-06, - "litellm_provider": "libertai", + "max_tokens": 262144, "mode": "chat", + "output_cost_per_token": 1.75e-06, + "source": "https://docs.libertai.io/apis/text/", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, - "supports_vision": true, "supports_reasoning": true, - "source": "https://docs.libertai.io/apis/text/" + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true }, "libertai/deepseek-v4-flash": { - "max_tokens": 200000, + "input_cost_per_token": 2.5e-07, + "litellm_provider": "libertai", "max_input_tokens": 200000, "max_output_tokens": 200000, - "input_cost_per_token": 2.5e-07, - "output_cost_per_token": 1.75e-06, - "litellm_provider": "libertai", + "max_tokens": 200000, "mode": "chat", + "output_cost_per_token": 1.75e-06, + "source": "https://docs.libertai.io/apis/text/", "supports_function_calling": true, - "supports_tool_choice": true, "supports_system_messages": true, - "supports_vision": false, - "source": "https://docs.libertai.io/apis/text/" + "supports_tool_choice": true, + "supports_vision": false }, "libertai/deepseek-v4-flash-thinking": { - "max_tokens": 200000, + "input_cost_per_token": 2.5e-07, + "litellm_provider": "libertai", "max_input_tokens": 200000, "max_output_tokens": 200000, - "input_cost_per_token": 2.5e-07, - "output_cost_per_token": 1.75e-06, - "litellm_provider": "libertai", + "max_tokens": 200000, "mode": "chat", + "output_cost_per_token": 1.75e-06, + "source": "https://docs.libertai.io/apis/text/", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, - "supports_vision": false, "supports_reasoning": true, - "source": "https://docs.libertai.io/apis/text/" + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": false }, "libertai/bge-m3": { - "max_tokens": 8192, - "max_input_tokens": 8192, "input_cost_per_token": 1e-08, - "output_cost_per_token": 0.0, "litellm_provider": "libertai", + "max_input_tokens": 8192, + "max_tokens": 8192, "mode": "embedding", + "output_cost_per_token": 0.0, "source": "https://docs.libertai.io/apis/text/" }, "sarvam/sarvam-m": { @@ -47916,6 +47920,8 @@ "search_context_size_medium": 0.01 }, "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -47924,9 +47930,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5-search-api-2025-10-14": { "cache_read_input_token_cost": 1.25e-07, @@ -47943,6 +47947,7 @@ "search_context_size_medium": 0.01 }, "supports_function_calling": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -47951,7 +47956,6 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, "supports_xhigh_reasoning_effort": false }, "gpt-realtime-mini-2025-10-06": { @@ -47992,6 +47996,7 @@ "cache_creation_input_audio_token_cost": 3e-07, "cache_read_input_audio_token_cost": 3e-07, "cache_read_input_token_cost": 6e-08, + "deprecation_date": "2027-01-20", "input_cost_per_audio_token": 1e-05, "input_cost_per_image": 8e-07, "input_cost_per_token": 6e-07, @@ -48019,8 +48024,7 @@ "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "deprecation_date": "2027-01-20" + "supports_tool_choice": true }, "gpt-realtime-whisper": { "input_cost_per_second": 0.0002833333333333333, @@ -48122,6 +48126,7 @@ "mode": "image_generation", "output_cost_per_image": 0.039, "output_cost_per_token": 0.0, + "rpm": 10, "source": "https://ai.google.dev/pricing", "supported_modalities": [ "text", @@ -48132,8 +48137,7 @@ "image" ], "supports_vision": true, - "tpm": 250000, - "rpm": 10 + "tpm": 250000 }, "gemini/gemini-2.0-flash-lite-001": { "cache_read_input_token_cost": 1.875e-08, @@ -48146,6 +48150,11 @@ "mode": "chat", "output_cost_per_token": 3e-07, "rpm": 4000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-2.0-flash-lite", "supported_modalities": [ "text", @@ -48164,14 +48173,10 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "tpm": 4000000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 4000000 }, "gemini-2.5-flash-native-audio-latest": { + "gemini_native_audio": true, "input_cost_per_audio_token": 1e-06, "input_cost_per_token": 3e-07, "litellm_provider": "gemini", @@ -48193,10 +48198,10 @@ "audio" ], "supports_audio_input": true, - "supports_audio_output": true, - "gemini_native_audio": true + "supports_audio_output": true }, "gemini-2.5-flash-native-audio-preview-09-2025": { + "gemini_native_audio": true, "input_cost_per_audio_token": 1e-06, "input_cost_per_token": 3e-07, "litellm_provider": "gemini", @@ -48218,10 +48223,10 @@ "audio" ], "supports_audio_input": true, - "supports_audio_output": true, - "gemini_native_audio": true + "supports_audio_output": true }, "gemini-2.5-flash-native-audio-preview-12-2025": { + "gemini_native_audio": true, "input_cost_per_audio_token": 3e-06, "input_cost_per_token": 5e-07, "litellm_provider": "gemini", @@ -48229,6 +48234,7 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "realtime", + "output_cost_per_audio_token": 1.2e-05, "output_cost_per_token": 2e-06, "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_endpoints": [ @@ -48243,11 +48249,10 @@ "audio" ], "supports_audio_input": true, - "supports_audio_output": true, - "gemini_native_audio": true, - "output_cost_per_audio_token": 1.2e-05 + "supports_audio_output": true }, "gemini-3.1-flash-live-preview": { + "gemini_audio_only_live": true, "input_cost_per_audio_token": 3e-06, "input_cost_per_image_token": 1e-06, "input_cost_per_token": 7.5e-07, @@ -48277,10 +48282,10 @@ "supports_audio_output": true, "supports_function_calling": true, "supports_vision": true, - "supports_web_search": true, - "gemini_audio_only_live": true + "supports_web_search": true }, "gemini/gemini-2.5-flash-native-audio-latest": { + "gemini_native_audio": true, "input_cost_per_audio_token": 1e-06, "input_cost_per_token": 3e-07, "litellm_provider": "gemini", @@ -48289,6 +48294,7 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 2.5e-06, + "rpm": 10, "source": "https://ai.google.dev/pricing", "supported_endpoints": [ "/v1/realtime" @@ -48303,11 +48309,10 @@ ], "supports_audio_input": true, "supports_audio_output": true, - "tpm": 250000, - "rpm": 10, - "gemini_native_audio": true + "tpm": 250000 }, "gemini/gemini-2.5-flash-native-audio-preview-09-2025": { + "gemini_native_audio": true, "input_cost_per_audio_token": 1e-06, "input_cost_per_token": 3e-07, "litellm_provider": "gemini", @@ -48316,6 +48321,7 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 2.5e-06, + "rpm": 10, "source": "https://ai.google.dev/pricing", "supported_endpoints": [ "/v1/realtime" @@ -48330,11 +48336,10 @@ ], "supports_audio_input": true, "supports_audio_output": true, - "tpm": 250000, - "rpm": 10, - "gemini_native_audio": true + "tpm": 250000 }, "gemini/gemini-2.5-flash-native-audio-preview-12-2025": { + "gemini_native_audio": true, "input_cost_per_audio_token": 3e-06, "input_cost_per_token": 5e-07, "litellm_provider": "gemini", @@ -48342,7 +48347,9 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "realtime", + "output_cost_per_audio_token": 1.2e-05, "output_cost_per_token": 2e-06, + "rpm": 10, "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_endpoints": [ "/v1/realtime" @@ -48357,12 +48364,10 @@ ], "supports_audio_input": true, "supports_audio_output": true, - "tpm": 250000, - "rpm": 10, - "gemini_native_audio": true, - "output_cost_per_audio_token": 1.2e-05 + "tpm": 250000 }, "gemini/gemini-3.1-flash-live-preview": { + "gemini_audio_only_live": true, "input_cost_per_audio_token": 3e-06, "input_cost_per_image_token": 1e-06, "input_cost_per_token": 7.5e-07, @@ -48374,6 +48379,7 @@ "mode": "realtime", "output_cost_per_audio_token": 1.2e-05, "output_cost_per_token": 4.5e-06, + "rpm": 10, "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_endpoints": [ "/v1/realtime" @@ -48393,9 +48399,7 @@ "supports_function_calling": true, "supports_vision": true, "supports_web_search": true, - "tpm": 250000, - "rpm": 10, - "gemini_audio_only_live": true + "tpm": 250000 }, "gemini/gemini-3.1-flash-tts-preview": { "input_cost_per_token": 1e-06, @@ -48405,12 +48409,12 @@ "max_tokens": 16384, "mode": "audio_speech", "output_cost_per_token": 2e-05, + "rpm": 10, "source": "https://ai.google.dev/gemini-api/docs/models/gemini-3.1-flash-tts-preview", "supported_endpoints": [ "/v1/audio/speech" ], - "tpm": 4000000, - "rpm": 10 + "tpm": 4000000 }, "gemini-2.5-flash-preview-tts": { "input_cost_per_token": 5e-07, @@ -48434,6 +48438,11 @@ "output_cost_per_reasoning_token": 2.5e-06, "output_cost_per_token": 2.5e-06, "rpm": 100000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-preview", "supported_endpoints": [ "/v1/chat/completions", @@ -48461,12 +48470,7 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 8000000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 8000000 }, "gemini-flash-lite-latest": { "cache_read_input_token_cost": 1e-08, @@ -48480,6 +48484,11 @@ "output_cost_per_reasoning_token": 4e-07, "output_cost_per_token": 4e-07, "rpm": 15, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-lite", "supported_endpoints": [ "/v1/chat/completions", @@ -48507,12 +48516,7 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 250000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 250000 }, "gemini-pro-latest": { "cache_read_input_token_cost": 1.25e-07, @@ -48527,6 +48531,11 @@ "output_cost_per_token": 1e-05, "output_cost_per_token_above_200k_tokens": 1.5e-05, "rpm": 2000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -48552,12 +48561,7 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "tpm": 800000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 800000 }, "gemini/gemini-pro-latest": { "cache_read_input_token_cost": 1.25e-07, @@ -48572,6 +48576,11 @@ "output_cost_per_token": 1e-05, "output_cost_per_token_above_200k_tokens": 1.5e-05, "rpm": 2000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -48597,12 +48606,7 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "tpm": 800000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 800000 }, "gemini-exp-1206": { "cache_read_input_token_cost": 3e-08, @@ -48616,6 +48620,11 @@ "output_cost_per_reasoning_token": 2.5e-06, "output_cost_per_token": 2.5e-06, "rpm": 100000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-preview", "supported_endpoints": [ "/v1/chat/completions", @@ -48643,20 +48652,13 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 8000000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 8000000 }, "vertex_ai/claude-sonnet-5@default": { - "deprecation_date": "2026-12-24", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 2.5e-06, "cache_creation_input_token_cost_above_1hr": 4e-06, "cache_read_input_token_cost": 2e-07, + "deprecation_date": "2026-12-24", "input_cost_per_token": 2e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -48664,6 +48666,8 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1e-05, + "prompt_cache_min_tokens": 1024, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -48673,6 +48677,8 @@ "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -48680,13 +48686,9 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "vertex_ai/claude-sonnet-4-6@default": { - "regional_endpoint_uplift_multiplier": 1.1, - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, "cache_read_input_token_cost": 3e-07, @@ -48697,40 +48699,43 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1.5e-05, - "supports_assistant_prefill": true, - "supports_computer_use": true, - "supports_function_calling": true, - "supports_pdf_input": true, - "supports_prompt_caching": true, - "supports_reasoning": true, - "supports_response_schema": true, - "supports_max_reasoning_effort": true, - "supports_tool_choice": true, - "supports_vision": true, + "prompt_cache_min_tokens": 1024, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, + "supports_assistant_prefill": true, + "supports_computer_use": true, + "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, "supports_output_config": true, - "prompt_cache_min_tokens": 1024 + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true }, "duckduckgo/search": { - "litellm_provider": "duckduckgo", - "mode": "search", "input_cost_per_query": 0.0, + "litellm_provider": "duckduckgo", "metadata": { "notes": "DuckDuckGo Instant Answer API is free and does not require an API key." - } + }, + "mode": "search" }, "bedrock_mantle/openai.gpt-oss-120b": { "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, "litellm_provider": "bedrock_mantle", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, "mode": "chat", + "output_cost_per_token": 6e-07, "supported_endpoints": [ "/v1/chat/completions", "/v1/responses" @@ -48743,12 +48748,12 @@ }, "bedrock_mantle/openai.gpt-oss-20b": { "input_cost_per_token": 7.5e-08, - "output_cost_per_token": 3e-07, "litellm_provider": "bedrock_mantle", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, "mode": "chat", + "output_cost_per_token": 3e-07, "supported_endpoints": [ "/v1/chat/completions", "/v1/responses" @@ -48761,12 +48766,12 @@ }, "bedrock_mantle/openai.gpt-oss-safeguard-120b": { "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, "litellm_provider": "bedrock_mantle", "max_input_tokens": 131072, "max_output_tokens": 65536, "max_tokens": 65536, "mode": "chat", + "output_cost_per_token": 6e-07, "supported_endpoints": [ "/v1/chat/completions" ], @@ -48777,12 +48782,12 @@ }, "bedrock_mantle/openai.gpt-oss-safeguard-20b": { "input_cost_per_token": 7.5e-08, - "output_cost_per_token": 3e-07, "litellm_provider": "bedrock_mantle", "max_input_tokens": 131072, "max_output_tokens": 65536, "max_tokens": 65536, "mode": "chat", + "output_cost_per_token": 3e-07, "supported_endpoints": [ "/v1/chat/completions" ], @@ -48792,20 +48797,19 @@ "supports_tool_choice": true }, "bedrock_mantle/openai.gpt-5.6-sol": { - "input_cost_per_token": 5.5e-06, - "input_cost_per_token_above_272k_tokens": 1.1e-05, "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_272k_tokens": 1.375e-05, "cache_read_input_token_cost": 5.5e-07, "cache_read_input_token_cost_above_272k_tokens": 1.1e-06, - "output_cost_per_token": 3.3e-05, - "output_cost_per_token_above_272k_tokens": 4.95e-05, + "input_cost_per_token": 5.5e-06, + "input_cost_per_token_above_272k_tokens": 1.1e-05, "litellm_provider": "bedrock_mantle", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "responses", - "use_openai_responses_path": true, + "output_cost_per_token": 3.3e-05, + "output_cost_per_token_above_272k_tokens": 4.95e-05, "supported_endpoints": [ "/v1/chat/completions", "/v1/responses" @@ -48818,6 +48822,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, + "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, @@ -48825,25 +48832,22 @@ "supports_vision": true, "supports_web_search": true, "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "supports_none_reasoning_effort": true, - "supports_native_streaming": true + "use_openai_responses_path": true }, "bedrock_mantle/openai.gpt-5.6-terra": { - "input_cost_per_token": 2.2e-06, - "input_cost_per_token_above_272k_tokens": 4.4e-06, "cache_creation_input_token_cost": 2.75e-06, "cache_creation_input_token_cost_above_272k_tokens": 5.5e-06, "cache_read_input_token_cost": 2.2e-07, "cache_read_input_token_cost_above_272k_tokens": 4.4e-07, - "output_cost_per_token": 1.32e-05, - "output_cost_per_token_above_272k_tokens": 1.98e-05, + "input_cost_per_token": 2.2e-06, + "input_cost_per_token_above_272k_tokens": 4.4e-06, "litellm_provider": "bedrock_mantle", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "responses", - "use_openai_responses_path": true, + "output_cost_per_token": 1.32e-05, + "output_cost_per_token_above_272k_tokens": 1.98e-05, "supported_endpoints": [ "/v1/chat/completions", "/v1/responses" @@ -48856,29 +48860,30 @@ "text" ], "supports_function_calling": true, + "supports_native_streaming": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_streaming": true + "use_openai_responses_path": true }, "bedrock_mantle/openai.gpt-5.6-luna": { - "input_cost_per_token": 2.2e-07, - "input_cost_per_token_above_272k_tokens": 4.4e-07, "cache_creation_input_token_cost": 2.75e-07, "cache_creation_input_token_cost_above_272k_tokens": 5.5e-07, "cache_read_input_token_cost": 2.2e-08, "cache_read_input_token_cost_above_272k_tokens": 4.4e-08, - "output_cost_per_token": 1.32e-06, - "output_cost_per_token_above_272k_tokens": 1.98e-06, + "input_cost_per_token": 2.2e-07, + "input_cost_per_token_above_272k_tokens": 4.4e-07, "litellm_provider": "bedrock_mantle", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "responses", - "use_openai_responses_path": true, + "output_cost_per_token": 1.32e-06, + "output_cost_per_token_above_272k_tokens": 1.98e-06, + "source": "https://aws.amazon.com/bedrock/", "supported_endpoints": [ "/v1/chat/completions", "/v1/responses" @@ -48891,29 +48896,29 @@ "text" ], "supports_function_calling": true, + "supports_native_streaming": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "source": "https://aws.amazon.com/bedrock/", "supports_web_search": true, - "supports_native_streaming": true + "use_openai_responses_path": true }, "us.openai.gpt-5.6-sol": { - "input_cost_per_token": 5.5e-06, - "input_cost_per_token_above_272k_tokens": 1.1e-05, "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_272k_tokens": 1.375e-05, "cache_read_input_token_cost": 5.5e-07, "cache_read_input_token_cost_above_272k_tokens": 1.1e-06, - "output_cost_per_token": 3.3e-05, - "output_cost_per_token_above_272k_tokens": 4.95e-05, + "input_cost_per_token": 5.5e-06, + "input_cost_per_token_above_272k_tokens": 1.1e-05, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.3e-05, + "output_cost_per_token_above_272k_tokens": 4.95e-05, "supported_modalities": [ "text", "image" @@ -48926,19 +48931,19 @@ "supports_vision": true }, "global.openai.gpt-5.6-sol": { - "input_cost_per_token": 5e-06, - "input_cost_per_token_above_272k_tokens": 1e-05, "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_272k_tokens": 1.25e-05, "cache_read_input_token_cost": 5e-07, "cache_read_input_token_cost_above_272k_tokens": 1e-06, - "output_cost_per_token": 3e-05, - "output_cost_per_token_above_272k_tokens": 4.5e-05, + "input_cost_per_token": 5e-06, + "input_cost_per_token_above_272k_tokens": 1e-05, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3e-05, + "output_cost_per_token_above_272k_tokens": 4.5e-05, "supported_modalities": [ "text", "image" @@ -48951,19 +48956,19 @@ "supports_vision": true }, "us.openai.gpt-5.6-terra": { - "input_cost_per_token": 2.2e-06, - "input_cost_per_token_above_272k_tokens": 4.4e-06, "cache_creation_input_token_cost": 2.75e-06, "cache_creation_input_token_cost_above_272k_tokens": 5.5e-06, "cache_read_input_token_cost": 2.2e-07, "cache_read_input_token_cost_above_272k_tokens": 4.4e-07, - "output_cost_per_token": 1.32e-05, - "output_cost_per_token_above_272k_tokens": 1.98e-05, + "input_cost_per_token": 2.2e-06, + "input_cost_per_token_above_272k_tokens": 4.4e-06, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1.32e-05, + "output_cost_per_token_above_272k_tokens": 1.98e-05, "supported_modalities": [ "text", "image" @@ -48976,19 +48981,19 @@ "supports_vision": true }, "global.openai.gpt-5.6-terra": { - "input_cost_per_token": 2e-06, - "input_cost_per_token_above_272k_tokens": 4e-06, "cache_creation_input_token_cost": 2.5e-06, "cache_creation_input_token_cost_above_272k_tokens": 5e-06, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_272k_tokens": 4e-07, - "output_cost_per_token": 1.2e-05, - "output_cost_per_token_above_272k_tokens": 1.8e-05, + "input_cost_per_token": 2e-06, + "input_cost_per_token_above_272k_tokens": 4e-06, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1.2e-05, + "output_cost_per_token_above_272k_tokens": 1.8e-05, "supported_modalities": [ "text", "image" @@ -49001,19 +49006,19 @@ "supports_vision": true }, "us.openai.gpt-5.6-luna": { - "input_cost_per_token": 2.2e-07, - "input_cost_per_token_above_272k_tokens": 4.4e-07, "cache_creation_input_token_cost": 2.75e-07, "cache_creation_input_token_cost_above_272k_tokens": 5.5e-07, "cache_read_input_token_cost": 2.2e-08, "cache_read_input_token_cost_above_272k_tokens": 4.4e-08, - "output_cost_per_token": 1.32e-06, - "output_cost_per_token_above_272k_tokens": 1.98e-06, + "input_cost_per_token": 2.2e-07, + "input_cost_per_token_above_272k_tokens": 4.4e-07, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1.32e-06, + "output_cost_per_token_above_272k_tokens": 1.98e-06, "supported_modalities": [ "text", "image" @@ -49026,19 +49031,19 @@ "supports_vision": true }, "global.openai.gpt-5.6-luna": { - "input_cost_per_token": 2e-07, - "input_cost_per_token_above_272k_tokens": 4e-07, "cache_creation_input_token_cost": 2.5e-07, "cache_creation_input_token_cost_above_272k_tokens": 5e-07, "cache_read_input_token_cost": 2e-08, "cache_read_input_token_cost_above_272k_tokens": 4e-08, - "output_cost_per_token": 1.2e-06, - "output_cost_per_token_above_272k_tokens": 1.8e-06, + "input_cost_per_token": 2e-07, + "input_cost_per_token_above_272k_tokens": 4e-07, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1.2e-06, + "output_cost_per_token_above_272k_tokens": 1.8e-06, "supported_modalities": [ "text", "image" @@ -49051,15 +49056,14 @@ "supports_vision": true }, "bedrock_mantle/openai.gpt-5.5": { - "input_cost_per_token": 5.5e-06, "cache_read_input_token_cost": 5.5e-07, - "output_cost_per_token": 3.3e-05, + "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_mantle", "max_input_tokens": 272000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "responses", - "use_openai_responses_path": true, + "output_cost_per_token": 3.3e-05, "supported_endpoints": [ "/v1/responses" ], @@ -49075,18 +49079,18 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true + "supports_vision": true, + "use_openai_responses_path": true }, "bedrock_mantle/openai.gpt-5.4": { - "input_cost_per_token": 2.75e-06, "cache_read_input_token_cost": 2.75e-07, - "output_cost_per_token": 1.65e-05, + "input_cost_per_token": 2.75e-06, "litellm_provider": "bedrock_mantle", "max_input_tokens": 272000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "responses", - "use_openai_responses_path": true, + "output_cost_per_token": 1.65e-05, "supported_endpoints": [ "/v1/responses" ], @@ -49102,17 +49106,17 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true + "supports_vision": true, + "use_openai_responses_path": true }, "bedrock_mantle/google.gemma-4-31b": { "input_cost_per_token": 1.4e-07, - "output_cost_per_token": 4e-07, "litellm_provider": "bedrock_mantle", "max_input_tokens": 256000, "max_output_tokens": 256000, "max_tokens": 256000, "mode": "chat", - "use_openai_responses_path": true, + "output_cost_per_token": 4e-07, "supported_endpoints": [ "/v1/chat/completions", "/v1/responses" @@ -49121,17 +49125,17 @@ "supports_parallel_function_calling": false, "supports_reasoning": true, "supports_tool_choice": true, - "supports_vision": true + "supports_vision": true, + "use_openai_responses_path": true }, "bedrock_mantle/google.gemma-4-26b-a4b": { "input_cost_per_token": 1.3e-07, - "output_cost_per_token": 4e-07, "litellm_provider": "bedrock_mantle", "max_input_tokens": 256000, "max_output_tokens": 256000, "max_tokens": 256000, "mode": "chat", - "use_openai_responses_path": true, + "output_cost_per_token": 4e-07, "supported_endpoints": [ "/v1/chat/completions", "/v1/responses" @@ -49140,17 +49144,17 @@ "supports_parallel_function_calling": false, "supports_reasoning": true, "supports_tool_choice": true, - "supports_vision": true + "supports_vision": true, + "use_openai_responses_path": true }, "bedrock_mantle/google.gemma-4-e2b": { "input_cost_per_token": 4e-08, - "output_cost_per_token": 8e-08, "litellm_provider": "bedrock_mantle", "max_input_tokens": 128000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", - "use_openai_responses_path": true, + "output_cost_per_token": 8e-08, "supported_endpoints": [ "/v1/chat/completions", "/v1/responses" @@ -49159,28 +49163,23 @@ "supports_parallel_function_calling": false, "supports_reasoning": true, "supports_tool_choice": true, - "supports_vision": true + "supports_vision": true, + "use_openai_responses_path": true }, "bedrock_mantle/xai.grok-4.3": { - "use_openai_responses_path": true, - "input_cost_per_token": 1.25e-06, - "output_cost_per_token": 2.5e-06, "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 1.25e-06, "litellm_provider": "bedrock_mantle", "max_input_tokens": 1000000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 2.5e-06, + "source": "https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-xai-grok-4-3.html", "supported_endpoints": [ "/v1/chat/completions", "/v1/responses" ], - "supports_function_calling": true, - "supports_reasoning": true, - "supports_response_schema": true, - "supports_tool_choice": true, - "supports_vision": true, - "source": "https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-xai-grok-4-3.html", "supported_modalities": [ "text", "image" @@ -49188,18 +49187,23 @@ "supported_output_modalities": [ "text" ], - "supports_none_reasoning_effort": true + "supports_function_calling": true, + "supports_none_reasoning_effort": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true, + "use_openai_responses_path": true }, "bedrock_mantle/xai.grok-4.6": { - "use_openai_responses_path": true, - "input_cost_per_token": 2.2e-06, - "output_cost_per_token": 6.6e-06, "cache_read_input_token_cost": 5.5e-07, + "input_cost_per_token": 2.2e-06, "litellm_provider": "bedrock_mantle", "max_input_tokens": 500000, "max_output_tokens": 500000, "max_tokens": 500000, "mode": "chat", + "output_cost_per_token": 6.6e-06, "supported_endpoints": [ "/v1/chat/completions", "/v1/responses" @@ -49209,17 +49213,18 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true + "supports_vision": true, + "use_openai_responses_path": true }, "us.xai.grok-4.6": { - "input_cost_per_token": 2.2e-06, - "output_cost_per_token": 6.6e-06, "cache_read_input_token_cost": 5.5e-07, + "input_cost_per_token": 2.2e-06, "litellm_provider": "bedrock_converse", "max_input_tokens": 500000, "max_output_tokens": 500000, "max_tokens": 500000, "mode": "chat", + "output_cost_per_token": 6.6e-06, "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -49227,14 +49232,14 @@ "supports_vision": true }, "global.xai.grok-4.6": { - "input_cost_per_token": 2e-06, - "output_cost_per_token": 6e-06, "cache_read_input_token_cost": 5e-07, + "input_cost_per_token": 2e-06, "litellm_provider": "bedrock_converse", "max_input_tokens": 500000, "max_output_tokens": 500000, "max_tokens": 500000, "mode": "chat", + "output_cost_per_token": 6e-06, "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -49395,33 +49400,33 @@ }, "bedrock/us-east-1/zai.glm-5": { "input_cost_per_token": 1e-06, - "output_cost_per_token": 3.2e-06, "litellm_provider": "bedrock", "max_input_tokens": 200000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, + "supports_native_structured_output": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_native_structured_output": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/us-west-2/zai.glm-5": { "input_cost_per_token": 1e-06, - "output_cost_per_token": 3.2e-06, "litellm_provider": "bedrock", "max_input_tokens": 200000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, + "supports_native_structured_output": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_native_structured_output": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/us-gov-east-1/anthropic.claude-haiku-4-5-20251001-v1:0": { "cache_creation_input_token_cost": 1.5e-06, @@ -49429,25 +49434,25 @@ "cache_read_input_token_cost": 1.2e-07, "input_cost_per_token": 1.2e-06, "litellm_provider": "bedrock", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 6e-06, + "prompt_cache_min_tokens": 4096, "source": "https://aws.amazon.com/about-aws/whats-new/2025/10/claude-4-5-haiku-anthropic-amazon-bedrock", "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, + "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_pdf_input": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "bedrock/us-gov-west-1/anthropic.claude-haiku-4-5-20251001-v1:0": { "cache_creation_input_token_cost": 1.5e-06, @@ -49455,211 +49460,212 @@ "cache_read_input_token_cost": 1.2e-07, "input_cost_per_token": 1.2e-06, "litellm_provider": "bedrock", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 6e-06, + "prompt_cache_min_tokens": 4096, "source": "https://aws.amazon.com/about-aws/whats-new/2025/10/claude-4-5-haiku-anthropic-amazon-bedrock", "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, + "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_pdf_input": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "snowflake/claude-sonnet-4-5": { - "max_tokens": 16384, + "cache_read_input_token_cost": 3e-07, + "input_cost_per_token": 3e-06, + "litellm_provider": "snowflake", "max_input_tokens": 200000, "max_output_tokens": 16384, - "input_cost_per_token": 3e-06, - "output_cost_per_token": 1.5e-05, - "cache_read_input_token_cost": 3e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 1.5e-05, + "prompt_cache_min_tokens": 1024, "supports_function_calling": true, - "supports_vision": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_response_schema": true, - "prompt_cache_min_tokens": 1024 + "supports_system_messages": true, + "supports_vision": true }, "snowflake/claude-sonnet-4-6": { - "supports_adaptive_thinking": true, - "max_tokens": 16384, + "cache_read_input_token_cost": 3e-07, + "input_cost_per_token": 3e-06, + "litellm_provider": "snowflake", "max_input_tokens": 200000, "max_output_tokens": 16384, - "input_cost_per_token": 3e-06, - "output_cost_per_token": 1.5e-05, - "cache_read_input_token_cost": 3e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 1.5e-05, + "prompt_cache_min_tokens": 1024, + "supports_adaptive_thinking": true, "supports_function_calling": true, - "supports_vision": true, + "supports_legacy_thinking": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_response_schema": true, - "prompt_cache_min_tokens": 1024 + "supports_system_messages": true, + "supports_vision": true }, "snowflake/claude-4-sonnet": { - "max_tokens": 16384, + "cache_read_input_token_cost": 3e-07, + "input_cost_per_token": 3e-06, + "litellm_provider": "snowflake", "max_input_tokens": 200000, "max_output_tokens": 16384, - "input_cost_per_token": 3e-06, - "output_cost_per_token": 1.5e-05, - "cache_read_input_token_cost": 3e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 1.5e-05, "supports_function_calling": true, - "supports_vision": true, "supports_prompt_caching": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_vision": true }, "snowflake/claude-4-opus": { - "max_tokens": 16384, + "cache_read_input_token_cost": 5e-07, + "input_cost_per_token": 5e-06, + "litellm_provider": "snowflake", "max_input_tokens": 200000, "max_output_tokens": 16384, - "input_cost_per_token": 5e-06, - "output_cost_per_token": 2.5e-05, - "cache_read_input_token_cost": 5e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 2.5e-05, "supports_function_calling": true, - "supports_vision": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_system_messages": true, + "supports_vision": true }, "snowflake/claude-haiku-4-5": { - "max_tokens": 16384, + "cache_read_input_token_cost": 1e-07, + "input_cost_per_token": 1e-06, + "litellm_provider": "snowflake", "max_input_tokens": 200000, "max_output_tokens": 16384, - "input_cost_per_token": 1e-06, - "output_cost_per_token": 5e-06, - "cache_read_input_token_cost": 1e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 5e-06, + "prompt_cache_min_tokens": 4096, "supports_function_calling": true, - "supports_vision": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_response_schema": true, - "prompt_cache_min_tokens": 4096 + "supports_system_messages": true, + "supports_vision": true }, "snowflake/claude-3-7-sonnet": { - "max_tokens": 16384, + "cache_read_input_token_cost": 3e-07, + "input_cost_per_token": 3e-06, + "litellm_provider": "snowflake", "max_input_tokens": 200000, "max_output_tokens": 16384, - "input_cost_per_token": 3e-06, - "output_cost_per_token": 1.5e-05, - "cache_read_input_token_cost": 3e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 1.5e-05, "supports_function_calling": true, - "supports_vision": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_system_messages": true, + "supports_vision": true }, "snowflake/openai-gpt-4.1": { - "max_tokens": 16384, + "cache_read_input_token_cost": 5e-07, + "input_cost_per_token": 2e-06, + "litellm_provider": "snowflake", "max_input_tokens": 300000, "max_output_tokens": 16384, - "input_cost_per_token": 2e-06, - "output_cost_per_token": 8e-06, - "cache_read_input_token_cost": 5e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 8e-06, "supports_function_calling": true, - "supports_vision": true, "supports_prompt_caching": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_vision": true }, "snowflake/openai-gpt-5": { - "max_tokens": 16384, + "cache_read_input_token_cost": 1.25e-07, + "input_cost_per_token": 1.25e-06, + "litellm_provider": "snowflake", "max_input_tokens": 300000, "max_output_tokens": 16384, - "input_cost_per_token": 1.25e-06, - "output_cost_per_token": 1e-05, - "cache_read_input_token_cost": 1.25e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 1e-05, "supports_function_calling": true, - "supports_vision": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_system_messages": true, + "supports_vision": true }, "snowflake/openai-gpt-5-mini": { - "max_tokens": 16384, + "input_cost_per_token": 3e-07, + "litellm_provider": "snowflake", "max_input_tokens": 1000000, "max_output_tokens": 16384, - "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 1.2e-06, "supports_function_calling": true, - "supports_system_messages": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_system_messages": true }, "snowflake/openai-gpt-5-nano": { - "max_tokens": 16384, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "snowflake", "max_input_tokens": 5000000, "max_output_tokens": 16384, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 6e-07, "supports_function_calling": true, - "supports_system_messages": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_system_messages": true }, "snowflake/llama4-maverick": { - "max_tokens": 16384, + "input_cost_per_token": 2.4e-07, + "litellm_provider": "snowflake", "max_input_tokens": 128000, "max_output_tokens": 16384, - "input_cost_per_token": 2.4e-07, - "output_cost_per_token": 9.7e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 9.7e-07, "supports_function_calling": true, "supports_system_messages": true }, "snowflake/snowflake-arctic-embed-l-v2.0": { - "max_tokens": 8192, - "max_input_tokens": 8192, "input_cost_per_token": 7e-08, - "output_cost_per_token": 0.0, "litellm_provider": "snowflake", - "mode": "embedding" + "max_input_tokens": 8192, + "max_tokens": 8192, + "mode": "embedding", + "output_cost_per_token": 0.0 }, "snowflake/snowflake-arctic-embed-m-v2.0": { - "max_tokens": 8192, - "max_input_tokens": 8192, "input_cost_per_token": 7e-08, - "output_cost_per_token": 0.0, "litellm_provider": "snowflake", - "mode": "embedding" + "max_input_tokens": 8192, + "max_tokens": 8192, + "mode": "embedding", + "output_cost_per_token": 0.0 }, "soniox/stt-async-v4": { + "input_cost_per_second": 0.0, "litellm_provider": "soniox", "max_output_tokens": 8000, "max_tokens": 8000, - "input_cost_per_second": 0.0, - "output_cost_per_second": 2.77778e-05, "mode": "audio_transcription", + "output_cost_per_second": 2.77778e-05, "source": "https://soniox.com/pricing", "supported_endpoints": [ "/v1/audio/transcriptions" @@ -49667,12 +49673,12 @@ "supports_audio_input": true }, "soniox/stt-async-v5": { + "input_cost_per_second": 0.0, "litellm_provider": "soniox", "max_output_tokens": 8000, "max_tokens": 8000, - "input_cost_per_second": 0.0, - "output_cost_per_second": 2.77778e-05, "mode": "audio_transcription", + "output_cost_per_second": 2.77778e-05, "source": "https://soniox.com/pricing", "supported_endpoints": [ "/v1/audio/transcriptions" @@ -49680,163 +49686,163 @@ "supports_audio_input": true }, "tensormesh/Qwen/Qwen3.5-397B-A17B-FP8": { - "litellm_provider": "tensormesh", - "mode": "chat", - "input_cost_per_token": 6e-07, - "output_cost_per_token": 3.6e-06, "cache_read_input_token_cost": 0, + "input_cost_per_token": 6e-07, + "litellm_provider": "tensormesh", "max_input_tokens": 262144, "max_output_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 3.6e-06, + "source": "https://serverless.tensormesh.ai/v1/models/openrouter", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "source": "https://serverless.tensormesh.ai/v1/models/openrouter" + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "tensormesh/Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8": { - "litellm_provider": "tensormesh", - "mode": "chat", - "input_cost_per_token": 4.5e-07, - "output_cost_per_token": 1.8e-06, "cache_read_input_token_cost": 0, + "input_cost_per_token": 4.5e-07, + "litellm_provider": "tensormesh", "max_input_tokens": 262144, "max_output_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 1.8e-06, + "source": "https://serverless.tensormesh.ai/v1/models/openrouter", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true, "supports_prompt_caching": true, + "supports_response_schema": true, "supports_system_messages": true, - "source": "https://serverless.tensormesh.ai/v1/models/openrouter" + "supports_tool_choice": true }, "tensormesh/Qwen/Qwen3.6-27B-FP8": { - "litellm_provider": "tensormesh", - "mode": "chat", - "input_cost_per_token": 3.2e-07, - "output_cost_per_token": 3.2e-06, "cache_read_input_token_cost": 0, + "input_cost_per_token": 3.2e-07, + "litellm_provider": "tensormesh", "max_input_tokens": 262144, "max_output_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 3.2e-06, + "source": "https://serverless.tensormesh.ai/v1/models/openrouter", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "source": "https://serverless.tensormesh.ai/v1/models/openrouter" + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "tensormesh/lukealonso/GLM-5.1-NVFP4-MTP": { - "litellm_provider": "tensormesh", - "mode": "chat", - "input_cost_per_token": 1.4e-06, - "output_cost_per_token": 4.4e-06, "cache_read_input_token_cost": 0, + "input_cost_per_token": 1.4e-06, + "litellm_provider": "tensormesh", "max_input_tokens": 202752, "max_output_tokens": 202752, + "mode": "chat", + "output_cost_per_token": 4.4e-06, + "source": "https://serverless.tensormesh.ai/v1/models/openrouter", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "source": "https://serverless.tensormesh.ai/v1/models/openrouter" + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "tensormesh/deepseek-ai/DeepSeek-V4-Flash": { - "litellm_provider": "tensormesh", - "mode": "chat", - "input_cost_per_token": 1.4e-07, - "output_cost_per_token": 2.8e-07, "cache_read_input_token_cost": 0, + "input_cost_per_token": 1.4e-07, + "litellm_provider": "tensormesh", "max_input_tokens": 32768, "max_output_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2.8e-07, + "source": "https://serverless.tensormesh.ai/v1/models/openrouter", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "source": "https://serverless.tensormesh.ai/v1/models/openrouter" + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "tensormesh/moonshotai/Kimi-K2.6": { - "litellm_provider": "tensormesh", - "mode": "chat", - "input_cost_per_token": 9.6e-07, - "output_cost_per_token": 4e-06, "cache_read_input_token_cost": 0, + "input_cost_per_token": 9.6e-07, + "litellm_provider": "tensormesh", "max_input_tokens": 32768, "max_output_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 4e-06, + "source": "https://serverless.tensormesh.ai/v1/models/openrouter", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "source": "https://serverless.tensormesh.ai/v1/models/openrouter" + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "tensormesh/MiniMaxAI/MiniMax-M2.5": { - "litellm_provider": "tensormesh", - "mode": "chat", - "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, "cache_read_input_token_cost": 0, + "input_cost_per_token": 3e-07, + "litellm_provider": "tensormesh", "max_input_tokens": 196608, "max_output_tokens": 196608, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://serverless.tensormesh.ai/v1/models/openrouter", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "source": "https://serverless.tensormesh.ai/v1/models/openrouter" + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "tensormesh/google/gemma-4-31B-it": { - "litellm_provider": "tensormesh", - "mode": "chat", - "input_cost_per_token": 1.4e-07, - "output_cost_per_token": 5.6e-07, "cache_read_input_token_cost": 0, + "input_cost_per_token": 1.4e-07, + "litellm_provider": "tensormesh", "max_input_tokens": 32768, "max_output_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 5.6e-07, + "source": "https://serverless.tensormesh.ai/v1/models/openrouter", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "source": "https://serverless.tensormesh.ai/v1/models/openrouter" + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "tensormesh/openai/gpt-oss-120b": { - "litellm_provider": "tensormesh", - "mode": "chat", - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, "cache_read_input_token_cost": 0, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "tensormesh", "max_input_tokens": 131072, "max_output_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 6e-07, + "source": "https://serverless.tensormesh.ai/v1/models/openrouter", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "source": "https://serverless.tensormesh.ai/v1/models/openrouter" + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "tensormesh/openai/gpt-oss-20b": { - "litellm_provider": "tensormesh", - "mode": "chat", - "input_cost_per_token": 7e-08, - "output_cost_per_token": 2.8e-07, "cache_read_input_token_cost": 0, + "input_cost_per_token": 7e-08, + "litellm_provider": "tensormesh", "max_input_tokens": 131072, "max_output_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2.8e-07, + "source": "https://serverless.tensormesh.ai/v1/models/openrouter", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "source": "https://serverless.tensormesh.ai/v1/models/openrouter" + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "deepseek-v4-flash": { "cache_creation_input_token_cost": 0.0, @@ -49962,10 +49968,10 @@ "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_prompt_caching": true, + "supports_reasoning": true, "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_reasoning": true, "supports_vision": false }, "tencent/deepseek-v4-flash": { @@ -49988,119 +49994,119 @@ "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_prompt_caching": true, + "supports_reasoning": true, "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_reasoning": true, "supports_vision": false }, "cognition/swe-1.6": { - "input_cost_per_token": 5e-07, - "output_cost_per_token": 2.5e-06, "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 5e-07, "litellm_provider": "cognition", "mode": "chat", + "output_cost_per_token": 2.5e-06, + "source": "https://docs.devin.ai/windsurf/plugins/cascade/models", "supports_function_calling": true, - "supports_prompt_caching": true, - "source": "https://docs.devin.ai/windsurf/plugins/cascade/models" + "supports_prompt_caching": true }, "cognition/swe-1.7": { - "input_cost_per_token": 5e-07, - "output_cost_per_token": 2.5e-06, "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 5e-07, "litellm_provider": "cognition", "mode": "chat", + "output_cost_per_token": 2.5e-06, + "source": "https://docs.devin.ai/desktop/models", "supports_function_calling": true, - "supports_prompt_caching": true, - "source": "https://docs.devin.ai/desktop/models" + "supports_prompt_caching": true }, "cognition/swe-1.7-lightning": { - "input_cost_per_token": 2.5e-06, - "output_cost_per_token": 1.25e-05, "cache_read_input_token_cost": 1e-06, + "input_cost_per_token": 2.5e-06, "litellm_provider": "cognition", "mode": "chat", + "output_cost_per_token": 1.25e-05, + "source": "https://docs.devin.ai/desktop/models", "supports_function_calling": true, - "supports_prompt_caching": true, - "source": "https://docs.devin.ai/desktop/models" + "supports_prompt_caching": true }, "pinstripes/ps/glm-4.5-air": { - "max_tokens": 128000, + "input_cost_per_token": 1.25e-07, + "litellm_provider": "pinstripes", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 1.25e-07, - "output_cost_per_token": 4.5e-07, - "litellm_provider": "pinstripes", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, + "output_cost_per_token": 4.5e-07, + "source": "https://pinstripes.io/pricing", "supports_assistant_prefill": true, - "supports_reasoning": true, - "source": "https://pinstripes.io/pricing" + "supports_function_calling": true, + "supports_reasoning": true }, "pinstripes/ps/qwen3.6-35b-a3b": { - "max_tokens": 131072, + "input_cost_per_token": 1.4e-07, + "litellm_provider": "pinstripes", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1.4e-07, - "output_cost_per_token": 4.5e-07, - "litellm_provider": "pinstripes", + "max_tokens": 131072, "mode": "chat", - "supports_function_calling": true, + "output_cost_per_token": 4.5e-07, + "source": "https://pinstripes.io/pricing", "supports_assistant_prefill": true, - "supports_reasoning": true, - "source": "https://pinstripes.io/pricing" + "supports_function_calling": true, + "supports_reasoning": true }, "pinstripes/ps/qwen3-30b-a3b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-08, + "litellm_provider": "pinstripes", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-08, - "output_cost_per_token": 2e-07, - "litellm_provider": "pinstripes", + "max_tokens": 131072, "mode": "chat", - "supports_function_calling": true, + "output_cost_per_token": 2e-07, + "source": "https://pinstripes.io/pricing", "supports_assistant_prefill": true, - "supports_reasoning": true, - "source": "https://pinstripes.io/pricing" + "supports_function_calling": true, + "supports_reasoning": true }, "pinstripes/ps/qwen3-coder-30b-a3b": { - "max_tokens": 131072, + "input_cost_per_token": 3e-07, + "litellm_provider": "pinstripes", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 3e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "pinstripes", + "max_tokens": 131072, "mode": "chat", - "supports_function_calling": true, + "output_cost_per_token": 6e-07, + "source": "https://pinstripes.io/pricing", "supports_assistant_prefill": true, - "supports_reasoning": false, - "source": "https://pinstripes.io/pricing" + "supports_function_calling": true, + "supports_reasoning": false }, "pinstripes/ps/deepseek-v4-flash": { - "max_tokens": 163840, + "input_cost_per_token": 1e-07, + "litellm_provider": "pinstripes", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "pinstripes", + "max_tokens": 163840, "mode": "chat", - "supports_function_calling": true, + "output_cost_per_token": 2e-07, + "source": "https://pinstripes.io/pricing", "supports_assistant_prefill": true, - "supports_reasoning": true, - "source": "https://pinstripes.io/pricing" + "supports_function_calling": true, + "supports_reasoning": true }, "pinstripes/ps/minimax-m2.7": { - "max_tokens": 1000192, + "input_cost_per_token": 2.55e-07, + "litellm_provider": "pinstripes", "max_input_tokens": 1000192, "max_output_tokens": 1000192, - "input_cost_per_token": 2.55e-07, - "output_cost_per_token": 5.5e-07, - "litellm_provider": "pinstripes", + "max_tokens": 1000192, "mode": "chat", - "supports_function_calling": true, + "output_cost_per_token": 5.5e-07, + "source": "https://pinstripes.io/pricing", "supports_assistant_prefill": true, - "supports_reasoning": false, - "source": "https://pinstripes.io/pricing" + "supports_function_calling": true, + "supports_reasoning": false }, "darkbloom/gemma-4-26b": { "input_cost_per_token": 3e-08, @@ -50138,63 +50144,63 @@ }, "xai/grok-4.20-0309-non-reasoning": { "cache_read_input_token_cost": 2e-07, + "cache_read_input_token_cost_above_200k_tokens": 4e-07, "input_cost_per_token": 1.25e-06, + "input_cost_per_token_above_200k_tokens": 2.5e-06, "litellm_provider": "xai", "max_input_tokens": 1000000, "max_output_tokens": 1000000, "max_tokens": 1000000, "mode": "chat", "output_cost_per_token": 2.5e-06, + "output_cost_per_token_above_200k_tokens": 5e-06, "source": "https://docs.x.ai/docs/models", "supports_function_calling": true, "supports_prompt_caching": true, + "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "input_cost_per_token_above_200k_tokens": 2.5e-06, - "output_cost_per_token_above_200k_tokens": 5e-06, - "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "supports_response_schema": true + "supports_web_search": true }, "xai/grok-4.20-multi-agent-0309": { "cache_read_input_token_cost": 2e-07, + "cache_read_input_token_cost_above_200k_tokens": 4e-07, "input_cost_per_token": 1.25e-06, + "input_cost_per_token_above_200k_tokens": 2.5e-06, "litellm_provider": "xai", "max_input_tokens": 1000000, "max_output_tokens": 1000000, "max_tokens": 1000000, "mode": "chat", "output_cost_per_token": 2.5e-06, + "output_cost_per_token_above_200k_tokens": 5e-06, "source": "https://docs.x.ai/docs/models", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, + "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "input_cost_per_token_above_200k_tokens": 2.5e-06, - "output_cost_per_token_above_200k_tokens": 5e-06, - "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "supports_response_schema": true + "supports_web_search": true }, "xai/grok-build-0.1": { "cache_read_input_token_cost": 2e-07, + "cache_read_input_token_cost_above_200k_tokens": 4e-07, "input_cost_per_token": 1e-06, + "input_cost_per_token_above_200k_tokens": 2e-06, "litellm_provider": "xai", "max_input_tokens": 256000, "max_output_tokens": 256000, "max_tokens": 256000, "mode": "chat", "output_cost_per_token": 2e-06, + "output_cost_per_token_above_200k_tokens": 4e-06, "source": "https://docs.x.ai/docs/models", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "input_cost_per_token_above_200k_tokens": 2e-06, - "output_cost_per_token_above_200k_tokens": 4e-06, - "cache_read_input_token_cost_above_200k_tokens": 4e-07, "supports_response_schema": true, + "supports_tool_choice": true, "supports_vision": true }, "gpt-transcribe": { @@ -50270,12 +50276,12 @@ }, "source": "https://docs.claude.com/en/docs/about-claude/models/overview", "supports_adaptive_thinking": true, - "thinking_always_on": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -50285,7 +50291,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true + "thinking_always_on": true }, "claude-mythos-preview": { "cache_creation_input_token_cost": 1.25e-05, @@ -50306,11 +50312,11 @@ }, "source": "https://docs.claude.com/en/docs/about-claude/models/overview", "supports_adaptive_thinking": true, - "thinking_always_on": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -50320,7 +50326,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true + "thinking_always_on": true }, "gemini/gemini-robotics-er-2-streaming-preview": { "input_cost_per_audio_token": 2e-06, @@ -50485,6 +50491,14 @@ "supports_adaptive_thinking": true } }, + { + "name": "claude-legacy-thinking", + "pattern": "claude-[a-z]+-4[-._]6(?!\\d)", + "description": "Claude at version 4.6 exactly, in any id shape that contains claude--4-6 (dotted and underscored minors included, dated releases such as claude-sonnet-4-6-20260219 too). The 4.6 family is adaptive-thinking yet still accepts legacy thinking.type=enabled with budget_tokens, so the caller's hard budget cap is forwarded verbatim instead of being rewritten to an uncapped output_config.effort. The lookahead keeps two-digit minors such as 4-60 from matching. 4.7+ and 5+ majors reject the legacy shape and stay on the adaptive translation.", + "model_info": { + "supports_legacy_thinking": true + } + }, { "name": "claude-always-on-thinking", "pattern": "claude-(?:fable|mythos)-", @@ -50922,6 +50936,42 @@ "supports_tool_choice": true, "supports_vision": true }, + "moonshot/kimi-k2.7-code": { + "cache_read_input_token_cost": 1.9e-07, + "input_cost_per_token": 9.5e-07, + "litellm_provider": "moonshot", + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 4e-06, + "source": "https://platform.kimi.ai/docs/pricing/chat-k27-code", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_video_input": true, + "supports_vision": true + }, + "moonshot/kimi-k2.7-code-highspeed": { + "cache_read_input_token_cost": 3.8e-07, + "input_cost_per_token": 1.9e-06, + "litellm_provider": "moonshot", + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 8e-06, + "source": "https://platform.kimi.ai/docs/pricing/chat-k27-code", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_video_input": true, + "supports_vision": true + }, "telnyx/moonshotai/Kimi-K2.6": { "litellm_provider": "telnyx", "max_input_tokens": 262144, diff --git a/litellm/proxy/video_endpoints/endpoints.py b/litellm/proxy/video_endpoints/endpoints.py index 6c6b004fd17..74688c58eaf 100644 --- a/litellm/proxy/video_endpoints/endpoints.py +++ b/litellm/proxy/video_endpoints/endpoints.py @@ -2,7 +2,6 @@ from typing import Any, Final -import orjson from fastapi import APIRouter, Depends, File, Form, Request, Response, UploadFile from fastapi.responses import ORJSONResponse @@ -20,6 +19,7 @@ from litellm.proxy.video_endpoints.utils import ( encode_character_id_in_response, extract_model_from_target_model_names, get_custom_provider_from_data, + video_reference_to_id, ) from litellm.types.videos.utils import ( decode_character_id_with_provider, @@ -451,9 +451,7 @@ async def video_remix( version, ) - # Read request body - body: Final = await request.body() - data: Final = orjson.loads(body) + data: Final = await _read_request_body(request=request) data["video_id"] = video_id decoded: Final = decode_video_id_with_provider(video_id) @@ -760,15 +758,10 @@ async def video_edit( version, ) - body: Final = await request.body() - data: Final = orjson.loads(body) + data: Final = await _read_request_body(request=request) + data["video_id"] = video_reference_to_id(data.pop("video", None)) - # Extract video_id from nested video object - video_ref: Final = data.pop("video", {}) - video_id: Final = video_ref.get("id", "") if isinstance(video_ref, dict) else "" - data["video_id"] = video_id - - decoded: Final = decode_video_id_with_provider(video_id) + decoded: Final = decode_video_id_with_provider(data["video_id"]) provider_from_id: Final = decoded.get("custom_llm_provider") model_id_from_decoded: Final = decoded.get("model_id") @@ -860,15 +853,10 @@ async def video_extension( version, ) - body: Final = await request.body() - data: Final = orjson.loads(body) + data: Final = await _read_request_body(request=request) + data["video_id"] = video_reference_to_id(data.pop("video", None)) - # Extract video_id from nested video object - video_ref: Final = data.pop("video", {}) - video_id: Final = video_ref.get("id", "") if isinstance(video_ref, dict) else "" - data["video_id"] = video_id - - decoded: Final = decode_video_id_with_provider(video_id) + decoded: Final = decode_video_id_with_provider(data["video_id"]) provider_from_id: Final = decoded.get("custom_llm_provider") model_id_from_decoded: Final = decoded.get("model_id") diff --git a/litellm/proxy/video_endpoints/utils.py b/litellm/proxy/video_endpoints/utils.py index d6b398e3476..a38226cc253 100644 --- a/litellm/proxy/video_endpoints/utils.py +++ b/litellm/proxy/video_endpoints/utils.py @@ -13,6 +13,18 @@ def extract_model_from_target_model_names(target_model_names: Any) -> str | None return target_model_names[0] if target_model_names else None +def video_reference_to_id(video_ref: object) -> str: + if isinstance(video_ref, dict): + return video_ref.get("id", "") + if not isinstance(video_ref, str): + return "" + try: + parsed_ref: Final = orjson.loads(video_ref) + except orjson.JSONDecodeError: + return video_ref + return parsed_ref.get("id", "") if isinstance(parsed_ref, dict) else video_ref + + def get_custom_provider_from_data(data: dict[str, Any]) -> str | None: custom_llm_provider: Final = data.get("custom_llm_provider") if custom_llm_provider: diff --git a/litellm/types/utils.py b/litellm/types/utils.py index 7da173db53c..d9d8170bff3 100644 --- a/litellm/types/utils.py +++ b/litellm/types/utils.py @@ -154,6 +154,7 @@ class ProviderSpecificModelInfo(TypedDict, total=False): supports_web_search: bool | None supports_reasoning: bool | None supports_adaptive_thinking: bool | None + supports_legacy_thinking: ReadOnly[bool | None] thinking_always_on: ReadOnly[bool | None] supports_tool_search: bool | None supports_mid_conversation_system: bool | None diff --git a/litellm/utils.py b/litellm/utils.py index e5ce7157e77..c0e26ef19d4 100644 --- a/litellm/utils.py +++ b/litellm/utils.py @@ -5753,6 +5753,7 @@ def _get_model_info_helper( supports_url_context=_model_info.get("supports_url_context", None), supports_reasoning=_model_info.get("supports_reasoning", None), supports_adaptive_thinking=_model_info.get("supports_adaptive_thinking", None), + supports_legacy_thinking=_model_info.get("supports_legacy_thinking", None), thinking_always_on=_model_info.get("thinking_always_on", None), supports_tool_search=_model_info.get("supports_tool_search", None), supports_mid_conversation_system=_model_info.get("supports_mid_conversation_system", None), diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index 79d9e579de7..f0ed171ad96 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -385,9 +385,9 @@ }, "amazon.nova-2-pro-preview-20251202-v1:0": { "cache_read_input_token_cost": 5.46875e-07, - "input_cost_per_token": 2.1875e-06, - "input_cost_per_image_token": 2.1875e-06, "input_cost_per_audio_token": 2.1875e-06, + "input_cost_per_image_token": 2.1875e-06, + "input_cost_per_token": 2.1875e-06, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 64000, @@ -404,9 +404,9 @@ }, "apac.amazon.nova-2-pro-preview-20251202-v1:0": { "cache_read_input_token_cost": 5.46875e-07, - "input_cost_per_token": 2.1875e-06, - "input_cost_per_image_token": 2.1875e-06, "input_cost_per_audio_token": 2.1875e-06, + "input_cost_per_image_token": 2.1875e-06, + "input_cost_per_token": 2.1875e-06, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 64000, @@ -440,9 +440,9 @@ }, "eu.amazon.nova-2-pro-preview-20251202-v1:0": { "cache_read_input_token_cost": 5.46875e-07, - "input_cost_per_token": 2.1875e-06, - "input_cost_per_image_token": 2.1875e-06, "input_cost_per_audio_token": 2.1875e-06, + "input_cost_per_image_token": 2.1875e-06, + "input_cost_per_token": 2.1875e-06, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 64000, @@ -476,9 +476,9 @@ }, "us.amazon.nova-2-pro-preview-20251202-v1:0": { "cache_read_input_token_cost": 5.46875e-07, - "input_cost_per_token": 2.1875e-06, - "input_cost_per_image_token": 2.1875e-06, "input_cost_per_audio_token": 2.1875e-06, + "input_cost_per_image_token": 2.1875e-06, + "input_cost_per_token": 2.1875e-06, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 64000, @@ -494,21 +494,21 @@ "supports_vision": true }, "amazon.nova-2-multimodal-embeddings-v1:0": { + "input_cost_per_audio_per_second": 0.00014, + "input_cost_per_image": 6e-05, + "input_cost_per_token": 1.35e-07, + "input_cost_per_video_per_second": 0.0007, "litellm_provider": "bedrock", "max_input_tokens": 8172, "max_tokens": 8172, "mode": "embedding", - "input_cost_per_token": 1.35e-07, - "input_cost_per_image": 6e-05, - "input_cost_per_video_per_second": 0.0007, - "input_cost_per_audio_per_second": 0.00014, "output_cost_per_token": 0.0, "output_vector_size": 3072, "source": "https://us-east-1.console.aws.amazon.com/bedrock/home?region=us-east-1#/model-catalog/serverless/amazon.nova-2-multimodal-embeddings-v1:0", + "supports_audio_input": true, "supports_embedding_image_input": true, "supports_image_input": true, - "supports_video_input": true, - "supports_audio_input": true + "supports_video_input": true }, "amazon.nova-micro-v1:0": { "input_cost_per_token": 3.5e-08, @@ -594,21 +594,21 @@ }, "amazon.titan-image-generator-v1": { "input_cost_per_image": 0.0, - "output_cost_per_image": 0.008, "litellm_provider": "bedrock", - "mode": "image_generation" + "mode": "image_generation", + "output_cost_per_image": 0.008 }, "amazon.titan-image-generator-v2": { "input_cost_per_image": 0.0, - "output_cost_per_image": 0.008, "litellm_provider": "bedrock", - "mode": "image_generation" + "mode": "image_generation", + "output_cost_per_image": 0.008 }, "amazon.titan-image-generator-v2:0": { "input_cost_per_image": 0.0, - "output_cost_per_image": 0.008, "litellm_provider": "bedrock", - "mode": "image_generation" + "mode": "image_generation", + "output_cost_per_image": 0.008 }, "twelvelabs.marengo-embed-2-7-v1:0": { "deprecation_date": "2026-11-30", @@ -624,10 +624,10 @@ }, "us.twelvelabs.marengo-embed-2-7-v1:0": { "deprecation_date": "2026-11-30", - "input_cost_per_token": 7e-05, - "input_cost_per_video_per_second": 0.0007, "input_cost_per_audio_per_second": 0.00014, "input_cost_per_image": 0.0001, + "input_cost_per_token": 7e-05, + "input_cost_per_video_per_second": 0.0007, "litellm_provider": "bedrock", "max_input_tokens": 77, "max_tokens": 77, @@ -639,10 +639,10 @@ }, "eu.twelvelabs.marengo-embed-2-7-v1:0": { "deprecation_date": "2026-11-30", - "input_cost_per_token": 7e-05, - "input_cost_per_video_per_second": 0.0007, "input_cost_per_audio_per_second": 0.00014, "input_cost_per_image": 0.0001, + "input_cost_per_token": 7e-05, + "input_cost_per_video_per_second": 0.0007, "litellm_provider": "bedrock", "max_input_tokens": 77, "max_tokens": 77, @@ -654,23 +654,23 @@ }, "twelvelabs.pegasus-1-2-v1:0": { "input_cost_per_video_per_second": 0.00049, - "output_cost_per_token": 7.5e-06, "litellm_provider": "bedrock", "mode": "chat", + "output_cost_per_token": 7.5e-06, "supports_video_input": true }, "us.twelvelabs.pegasus-1-2-v1:0": { "input_cost_per_video_per_second": 0.00049, - "output_cost_per_token": 7.5e-06, "litellm_provider": "bedrock", "mode": "chat", + "output_cost_per_token": 7.5e-06, "supports_video_input": true }, "eu.twelvelabs.pegasus-1-2-v1:0": { "input_cost_per_video_per_second": 0.00049, - "output_cost_per_token": 7.5e-06, "litellm_provider": "bedrock", "mode": "chat", + "output_cost_per_token": 7.5e-06, "supports_video_input": true }, "amazon.titan-text-express-v1": { @@ -710,41 +710,41 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 4e-06, + "prompt_cache_min_tokens": 2048, "supports_assistant_prefill": true, "supports_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_response_schema": true, - "supports_tool_choice": true, - "prompt_cache_min_tokens": 2048 + "supports_tool_choice": true }, "anthropic.claude-haiku-4-5-20251001-v1:0": { "cache_creation_input_token_cost": 1.25e-06, "cache_creation_input_token_cost_above_1hr": 2e-06, "cache_read_input_token_cost": 1e-07, "input_cost_per_token": 1e-06, + "input_cost_per_token_batches": 5e-07, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, + "output_cost_per_token_batches": 2.5e-06, + "prompt_cache_min_tokens": 4096, "source": "https://aws.amazon.com/about-aws/whats-new/2025/10/claude-4-5-haiku-anthropic-amazon-bedrock", "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096, - "input_cost_per_token_batches": 5e-07, - "output_cost_per_token_batches": 2.5e-06 + "supports_tool_search": true, + "supports_vision": true }, "anthropic.claude-haiku-4-5@20251001": { "cache_creation_input_token_cost": 1.25e-06, @@ -752,28 +752,30 @@ "cache_read_input_token_cost": 1e-07, "input_cost_per_token": 1e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, + "prompt_cache_min_tokens": 4096, "source": "https://aws.amazon.com/about-aws/whats-new/2025/10/claude-4-5-haiku-anthropic-amazon-bedrock", "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_streaming": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_streaming": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "anthropic.claude-3-5-sonnet-20240620-v1:0": { + "cache_creation_input_token_cost": 3.75e-06, + "cache_read_input_token_cost": 3e-07, "input_cost_per_token": 3e-06, "litellm_provider": "bedrock", "max_input_tokens": 200000, @@ -785,9 +787,7 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_creation_input_token_cost": 3.75e-06, - "cache_read_input_token_cost": 3e-07 + "supports_vision": true }, "anthropic.claude-3-5-sonnet-20241022-v2:0": { "cache_creation_input_token_cost": 3.75e-06, @@ -849,6 +849,8 @@ "supports_vision": true }, "anthropic.claude-3-haiku-20240307-v1:0": { + "cache_creation_input_token_cost": 3.125e-07, + "cache_read_input_token_cost": 2.5e-08, "deprecation_date": "2026-09-10", "input_cost_per_token": 2.5e-07, "litellm_provider": "bedrock", @@ -861,11 +863,11 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 2.5e-08, - "cache_creation_input_token_cost": 3.125e-07 + "supports_vision": true }, "anthropic.claude-3-opus-20240229-v1:0": { + "cache_creation_input_token_cost": 1.875e-05, + "cache_read_input_token_cost": 1.5e-06, "input_cost_per_token": 1.5e-05, "litellm_provider": "bedrock", "max_input_tokens": 200000, @@ -876,11 +878,11 @@ "supports_function_calling": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 1.5e-06, - "cache_creation_input_token_cost": 1.875e-05 + "supports_vision": true }, "anthropic.claude-3-sonnet-20240229-v1:0": { + "cache_creation_input_token_cost": 3.75e-06, + "cache_read_input_token_cost": 3e-07, "deprecation_date": "2026-07-30", "input_cost_per_token": 3e-06, "litellm_provider": "bedrock", @@ -893,9 +895,7 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-07, - "cache_creation_input_token_cost": 3.75e-06 + "supports_vision": true }, "anthropic.claude-instant-v1": { "input_cost_per_token": 8e-07, @@ -918,6 +918,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -931,8 +932,7 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "anthropic.claude-opus-4-20250514-v1:0": { "cache_creation_input_token_cost": 1.875e-05, @@ -944,6 +944,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -957,21 +958,21 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "anthropic.claude-opus-4-5-20251101-v1:0": { + "bedrock_output_config_effort_ceiling": "high", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -980,390 +981,395 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "high", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "anthropic.claude-opus-4-6-v1": { - "supports_adaptive_thinking": true, + "bedrock_output_config_effort_ceiling": "max", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "supports_max_reasoning_effort": true, - "bedrock_output_config_effort_ceiling": "max", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "global.anthropic.claude-opus-4-6-v1": { - "supports_adaptive_thinking": true, + "bedrock_output_config_effort_ceiling": "max", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "supports_max_reasoning_effort": true, - "bedrock_output_config_effort_ceiling": "max", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "us.anthropic.claude-opus-4-6-v1": { - "supports_adaptive_thinking": true, + "bedrock_output_config_effort_ceiling": "max", "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "supports_max_reasoning_effort": true, - "bedrock_output_config_effort_ceiling": "max", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "eu.anthropic.claude-opus-4-6-v1": { - "supports_adaptive_thinking": true, + "bedrock_output_config_effort_ceiling": "max", "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "supports_max_reasoning_effort": true, - "bedrock_output_config_effort_ceiling": "max", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "au.anthropic.claude-opus-4-6-v1": { - "supports_adaptive_thinking": true, + "bedrock_output_config_effort_ceiling": "max", "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "supports_max_reasoning_effort": true, - "bedrock_output_config_effort_ceiling": "max", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "anthropic.claude-opus-4-7": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 2048, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_sampling_params": false, "supports_tool_choice": true, + "supports_tool_search": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "anthropic.claude-mythos-preview": { "input_cost_per_token": 0, - "output_cost_per_token": 0, "litellm_provider": "bedrock", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", - "thinking_always_on": true, + "output_cost_per_token": 0, "supports_function_calling": true, - "supports_vision": true, + "supports_output_config": true, "supports_prompt_caching": false, "supports_reasoning": true, "supports_tool_choice": true, - "supports_output_config": true + "supports_vision": true, + "thinking_always_on": true }, "global.anthropic.claude-opus-4-7": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 2048, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_sampling_params": false, "supports_tool_choice": true, + "supports_tool_search": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "us.anthropic.claude-opus-4-7": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 2048, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_sampling_params": false, "supports_tool_choice": true, + "supports_tool_search": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "eu.anthropic.claude-opus-4-7": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 2048, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_sampling_params": false, "supports_tool_choice": true, + "supports_tool_search": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "au.anthropic.claude-opus-4-7": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 2048, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_sampling_params": false, "supports_tool_choice": true, + "supports_tool_search": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "anthropic.claude-fable-5": { + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 1.25e-05, "cache_creation_input_token_cost_above_1hr": 2e-05, "cache_read_input_token_cost": 1e-06, @@ -1374,17 +1380,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 5e-05, + "prompt_cache_min_tokens": 512, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "thinking_always_on": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1393,14 +1403,10 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 512 + "thinking_always_on": true }, "global.anthropic.claude-fable-5": { + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 1.25e-05, "cache_creation_input_token_cost_above_1hr": 2e-05, "cache_read_input_token_cost": 1e-06, @@ -1411,17 +1417,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 5e-05, + "prompt_cache_min_tokens": 512, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "thinking_always_on": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1430,14 +1440,10 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 512 + "thinking_always_on": true }, "us.anthropic.claude-fable-5": { + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 1.375e-05, "cache_creation_input_token_cost_above_1hr": 2.2e-05, "cache_read_input_token_cost": 1.1e-06, @@ -1448,17 +1454,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 5.5e-05, + "prompt_cache_min_tokens": 512, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "thinking_always_on": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1467,14 +1477,10 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 512 + "thinking_always_on": true }, "eu.anthropic.claude-fable-5": { + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 1.375e-05, "cache_creation_input_token_cost_above_1hr": 2.2e-05, "cache_read_input_token_cost": 1.1e-06, @@ -1485,17 +1491,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 5.5e-05, + "prompt_cache_min_tokens": 512, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "thinking_always_on": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1504,17 +1514,10 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 512 + "thinking_always_on": true }, "anthropic.claude-opus-5": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, @@ -1525,14 +1528,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 512, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1540,17 +1550,10 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 512 + "supports_xhigh_reasoning_effort": true }, "global.anthropic.claude-opus-5": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, @@ -1561,14 +1564,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 512, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1576,17 +1586,10 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 512 + "supports_xhigh_reasoning_effort": true }, "us.anthropic.claude-opus-5": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, @@ -1597,14 +1600,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 512, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1612,17 +1622,10 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 512 + "supports_xhigh_reasoning_effort": true }, "eu.anthropic.claude-opus-5": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, @@ -1633,14 +1636,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 512, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1648,17 +1658,10 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 512 + "supports_xhigh_reasoning_effort": true }, "au.anthropic.claude-opus-5": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, @@ -1669,14 +1672,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 512, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1684,17 +1694,10 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 512 + "supports_xhigh_reasoning_effort": true }, "jp.anthropic.claude-opus-5": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, @@ -1705,14 +1708,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 512, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1720,17 +1730,11 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 512 + "supports_xhigh_reasoning_effort": true }, "anthropic.claude-opus-4-8": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, @@ -1741,14 +1745,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1756,18 +1767,11 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "global.anthropic.claude-opus-4-8": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, @@ -1778,14 +1782,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1793,18 +1804,11 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "us.anthropic.claude-opus-4-8": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, @@ -1815,14 +1819,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1830,18 +1841,11 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "eu.anthropic.claude-opus-4-8": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, @@ -1852,14 +1856,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1867,18 +1878,11 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "au.anthropic.claude-opus-4-8": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, @@ -1889,14 +1893,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1904,18 +1915,11 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "jp.anthropic.claude-opus-4-8": { "bedrock_converse_supports_strict_tools": false, - "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, @@ -1926,14 +1930,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -1941,13 +1952,7 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "jp.anthropic.claude-opus-4-7": { "bedrock_converse_supports_strict_tools": false, @@ -1955,12 +1960,12 @@ "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 2048, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -1970,22 +1975,23 @@ "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_minimal_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_sampling_params": false, "supports_tool_choice": true, + "supports_tool_search": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_minimal_reasoning_effort": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "anthropic.claude-sonnet-5": { "bedrock_converse_supports_strict_tools": false, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 2.5e-06, "cache_creation_input_token_cost_above_1hr": 4e-06, "cache_read_input_token_cost": 2e-07, @@ -1996,16 +2002,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -2013,16 +2024,11 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "global.anthropic.claude-sonnet-5": { "bedrock_converse_supports_strict_tools": false, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 2.5e-06, "cache_creation_input_token_cost_above_1hr": 4e-06, "cache_read_input_token_cost": 2e-07, @@ -2033,16 +2039,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -2050,16 +2061,11 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "us.anthropic.claude-sonnet-5": { "bedrock_converse_supports_strict_tools": false, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 2.75e-06, "cache_creation_input_token_cost_above_1hr": 4.4e-06, "cache_read_input_token_cost": 2.2e-07, @@ -2070,16 +2076,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1.1e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -2087,16 +2098,11 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "eu.anthropic.claude-sonnet-5": { "bedrock_converse_supports_strict_tools": false, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 2.75e-06, "cache_creation_input_token_cost_above_1hr": 4.4e-06, "cache_read_input_token_cost": 2.2e-07, @@ -2107,16 +2113,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1.1e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -2124,16 +2135,11 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "au.anthropic.claude-sonnet-5": { "bedrock_converse_supports_strict_tools": false, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 2.75e-06, "cache_creation_input_token_cost_above_1hr": 4.4e-06, "cache_read_input_token_cost": 2.2e-07, @@ -2144,16 +2150,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1.1e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -2161,16 +2172,11 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "jp.anthropic.claude-sonnet-5": { "bedrock_converse_supports_strict_tools": false, + "bedrock_output_config_effort_ceiling": "xhigh", "cache_creation_input_token_cost": 2.75e-06, "cache_creation_input_token_cost_above_1hr": 4.4e-06, "cache_read_input_token_cost": 2.2e-07, @@ -2181,16 +2187,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1.1e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -2198,227 +2209,229 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "xhigh", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "anthropic.claude-sonnet-4-6": { - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, "cache_read_input_token_cost": 3e-07, "input_cost_per_token": 3e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_tool_search": true, + "supports_vision": true }, "global.anthropic.claude-sonnet-4-6": { - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, "cache_read_input_token_cost": 3e-07, "input_cost_per_token": 3e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_tool_search": true, + "supports_vision": true }, "us.anthropic.claude-sonnet-4-6": { - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 4.125e-06, "cache_creation_input_token_cost_above_1hr": 6.6e-06, "cache_read_input_token_cost": 3.3e-07, "input_cost_per_token": 3.3e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.65e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_tool_search": true, + "supports_vision": true }, "eu.anthropic.claude-sonnet-4-6": { - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 4.125e-06, "cache_creation_input_token_cost_above_1hr": 6.6e-06, "cache_read_input_token_cost": 3.3e-07, "input_cost_per_token": 3.3e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.65e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_tool_search": true, + "supports_vision": true }, "au.anthropic.claude-sonnet-4-6": { - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 4.125e-06, "cache_creation_input_token_cost_above_1hr": 6.6e-06, "cache_read_input_token_cost": 3.3e-07, "input_cost_per_token": 3.3e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.65e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_tool_search": true, + "supports_vision": true }, "jp.anthropic.claude-sonnet-4-6": { - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 4.125e-06, "cache_creation_input_token_cost_above_1hr": 6.6e-06, "cache_read_input_token_cost": 3.3e-07, "input_cost_per_token": 3.3e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.65e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_tool_search": true, + "supports_vision": true }, "anthropic.claude-sonnet-4-20250514-v1:0": { + "bedrock_converse_supports_strict_tools": false, "cache_creation_input_token_cost": 3.75e-06, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, "deprecation_date": "2026-10-14", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -2432,27 +2445,27 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "bedrock_converse_supports_strict_tools": false, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "anthropic.claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, + "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.2e-05, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.2e-05, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "input_cost_per_token_batches": 1.5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "output_cost_per_token_batches": 7.5e-06, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -2461,17 +2474,15 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024, - "input_cost_per_token_batches": 1.5e-06, - "output_cost_per_token_batches": 7.5e-06 + "supports_tool_search": true, + "supports_vision": true }, "anthropic.claude-v1": { "input_cost_per_token": 8e-06, @@ -2651,6 +2662,8 @@ "supports_vision": true }, "apac.anthropic.claude-3-5-sonnet-20240620-v1:0": { + "cache_creation_input_token_cost": 3.75e-06, + "cache_read_input_token_cost": 3e-07, "deprecation_date": "2026-07-30", "input_cost_per_token": 3e-06, "litellm_provider": "bedrock", @@ -2663,9 +2676,7 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-07, - "cache_creation_input_token_cost": 3.75e-06 + "supports_vision": true }, "apac.anthropic.claude-3-5-sonnet-20241022-v2:0": { "cache_creation_input_token_cost": 3.75e-06, @@ -2688,6 +2699,8 @@ "supports_vision": true }, "apac.anthropic.claude-3-haiku-20240307-v1:0": { + "cache_creation_input_token_cost": 3.125e-07, + "cache_read_input_token_cost": 2.5e-08, "deprecation_date": "2026-09-10", "input_cost_per_token": 2.5e-07, "litellm_provider": "bedrock", @@ -2700,38 +2713,38 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 2.5e-08, - "cache_creation_input_token_cost": 3.125e-07 + "supports_vision": true }, "apac.anthropic.claude-haiku-4-5-20251001-v1:0": { "cache_creation_input_token_cost": 1.375e-06, "cache_read_input_token_cost": 1.1e-07, "input_cost_per_token": 1.1e-06, + "input_cost_per_token_batches": 5.5e-07, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5.5e-06, + "output_cost_per_token_batches": 2.75e-06, + "prompt_cache_min_tokens": 4096, "source": "https://aws.amazon.com/about-aws/whats-new/2025/10/claude-4-5-haiku-anthropic-amazon-bedrock", "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096, - "input_cost_per_token_batches": 5.5e-07, - "output_cost_per_token_batches": 2.75e-06 + "supports_tool_search": true, + "supports_vision": true }, "apac.anthropic.claude-3-sonnet-20240229-v1:0": { + "cache_creation_input_token_cost": 3.75e-06, + "cache_read_input_token_cost": 3e-07, "deprecation_date": "2026-07-30", "input_cost_per_token": 3e-06, "litellm_provider": "bedrock", @@ -2744,25 +2757,25 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-07, - "cache_creation_input_token_cost": 3.75e-06 + "supports_vision": true }, "apac.anthropic.claude-sonnet-4-20250514-v1:0": { + "bedrock_converse_supports_strict_tools": false, "cache_creation_input_token_cost": 3.75e-06, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, "deprecation_date": "2026-10-14", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -2776,9 +2789,7 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "bedrock_converse_supports_strict_tools": false, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "assemblyai/best": { "input_cost_per_second": 3.333e-05, @@ -2795,20 +2806,22 @@ "au.anthropic.claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 4.125e-06, "cache_creation_input_token_cost_above_1hr": 6.6e-06, + "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.32e-05, + "cache_creation_input_token_cost_above_200k_tokens": 8.25e-06, "cache_read_input_token_cost": 3.3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6.6e-07, "input_cost_per_token": 3.3e-06, "input_cost_per_token_above_200k_tokens": 6.6e-06, - "output_cost_per_token_above_200k_tokens": 2.475e-05, - "cache_creation_input_token_cost_above_200k_tokens": 8.25e-06, - "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.32e-05, - "cache_read_input_token_cost_above_200k_tokens": 6.6e-07, + "input_cost_per_token_batches": 1.65e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.65e-05, + "output_cost_per_token_above_200k_tokens": 2.475e-05, + "output_cost_per_token_batches": 8.25e-06, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -2817,17 +2830,15 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024, - "input_cost_per_token_batches": 1.65e-06, - "output_cost_per_token_batches": 8.25e-06 + "supports_tool_search": true, + "supports_vision": true }, "azure/ada": { "input_cost_per_token": 1e-07, @@ -2878,10 +2889,10 @@ "supports_function_calling": true }, "azure_ai/claude-haiku-4-5": { - "deprecation_date": "2026-10-19", "cache_creation_input_token_cost": 1.25e-06, "cache_creation_input_token_cost_above_1hr": 2e-06, "cache_read_input_token_cost": 1e-07, + "deprecation_date": "2026-10-19", "input_cost_per_token": 1e-06, "litellm_provider": "azure_ai", "max_input_tokens": 200000, @@ -2889,6 +2900,7 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, + "prompt_cache_min_tokens": 4096, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, @@ -2897,14 +2909,13 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "azure_ai/claude-opus-4-5": { - "deprecation_date": "2026-10-19", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2026-10-19", "input_cost_per_token": 5e-06, "litellm_provider": "azure_ai", "max_input_tokens": 200000, @@ -2912,70 +2923,73 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "azure_ai/claude-opus-4-6": { + "cache_creation_input_token_cost": 6.25e-06, + "cache_creation_input_token_cost_above_1hr": 1e-05, + "cache_read_input_token_cost": 5e-07, "deprecation_date": "2027-02-02", - "supports_adaptive_thinking": true, "input_cost_per_token": 5e-06, - "output_cost_per_token": 2.5e-05, "litellm_provider": "azure_ai", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, - "cache_creation_input_token_cost": 6.25e-06, - "cache_creation_input_token_cost_above_1hr": 1e-05, - "cache_read_input_token_cost": 5e-07, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "azure_ai/claude-opus-4-7": { + "cache_creation_input_token_cost": 6.25e-06, + "cache_creation_input_token_cost_above_1hr": 1e-05, + "cache_read_input_token_cost": 5e-07, "deprecation_date": "2027-04-06", - "supports_adaptive_thinking": true, "input_cost_per_token": 5e-06, - "output_cost_per_token": 2.5e-05, "litellm_provider": "azure_ai", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 2048, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, - "cache_creation_input_token_cost": 6.25e-06, - "cache_creation_input_token_cost_above_1hr": 1e-05, - "cache_read_input_token_cost": 5e-07, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -2983,32 +2997,31 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "azure_ai/claude-fable-5": { - "supports_mid_conversation_system": true, - "input_cost_per_token": 1e-05, - "output_cost_per_token": 5e-05, - "litellm_provider": "azure_ai", - "max_input_tokens": 1000000, - "max_output_tokens": 128000, - "max_tokens": 128000, - "mode": "chat", - "search_context_cost_per_query": { - "search_context_size_high": 0.01, - "search_context_size_low": 0.01, - "search_context_size_medium": 0.01 - }, "cache_creation_input_token_cost": 1.25e-05, "cache_creation_input_token_cost_above_1hr": 2e-05, "cache_read_input_token_cost": 1e-06, + "input_cost_per_token": 1e-05, + "litellm_provider": "azure_ai", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 5e-05, + "prompt_cache_min_tokens": 512, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, "supports_adaptive_thinking": true, - "thinking_always_on": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -3017,30 +3030,31 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 512 + "thinking_always_on": true }, "azure_ai/claude-opus-5": { - "supports_mid_conversation_system": true, - "supports_adaptive_thinking": true, + "cache_creation_input_token_cost": 6.25e-06, + "cache_creation_input_token_cost_above_1hr": 1e-05, + "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, - "output_cost_per_token": 2.5e-05, "litellm_provider": "azure_ai", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 512, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, - "cache_creation_input_token_cost": 6.25e-06, - "cache_creation_input_token_cost_above_1hr": 1e-05, - "cache_read_input_token_cost": 5e-07, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -3048,31 +3062,31 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 512 + "supports_xhigh_reasoning_effort": true }, "azure_ai/claude-opus-4-8": { - "supports_mid_conversation_system": true, - "supports_adaptive_thinking": true, + "cache_creation_input_token_cost": 6.25e-06, + "cache_creation_input_token_cost_above_1hr": 1e-05, + "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, - "output_cost_per_token": 2.5e-05, "litellm_provider": "azure_ai", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, - "cache_creation_input_token_cost": 6.25e-06, - "cache_creation_input_token_cost_above_1hr": 1e-05, - "cache_read_input_token_cost": 5e-07, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -3080,15 +3094,13 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "azure_ai/claude-opus-4-1": { - "deprecation_date": "2026-08-05", "cache_creation_input_token_cost": 1.875e-05, "cache_creation_input_token_cost_above_1hr": 3e-05, "cache_read_input_token_cost": 1.5e-06, + "deprecation_date": "2026-08-05", "input_cost_per_token": 1.5e-05, "litellm_provider": "azure_ai", "max_input_tokens": 200000, @@ -3096,6 +3108,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, @@ -3104,14 +3117,13 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "azure_ai/claude-sonnet-4-5": { - "deprecation_date": "2026-10-19", "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, "cache_read_input_token_cost": 3e-07, + "deprecation_date": "2026-10-19", "input_cost_per_token": 3e-06, "litellm_provider": "azure_ai", "max_input_tokens": 200000, @@ -3119,6 +3131,7 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, @@ -3127,11 +3140,9 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "azure_ai/claude-sonnet-5": { - "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 2.5e-06, "cache_creation_input_token_cost_above_1hr": 4e-06, "cache_read_input_token_cost": 2e-07, @@ -3142,6 +3153,7 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -3151,6 +3163,8 @@ "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -3158,16 +3172,13 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "azure_ai/claude-sonnet-4-6": { - "deprecation_date": "2027-02-10", - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, "cache_read_input_token_cost": 3e-07, + "deprecation_date": "2027-02-10", "input_cost_per_token": 3e-06, "litellm_provider": "azure_ai", "max_input_tokens": 1000000, @@ -3175,18 +3186,20 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "prompt_cache_min_tokens": 1024, + "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "azure/computer-use-preview": { "input_cost_per_token": 3e-06, @@ -3222,12 +3235,12 @@ }, "azure_ai/gpt-oss-120b": { "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, "litellm_provider": "azure_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 6e-07, "source": "https://azure.microsoft.com/en-us/pricing/details/cognitive-services/openai-service/", "supports_function_calling": true, "supports_parallel_function_calling": true, @@ -3235,15 +3248,15 @@ "supports_tool_choice": true }, "azure_ai/gpt-5.5": { - "deprecation_date": "2027-10-26", "cache_read_input_token_cost": 5e-07, "cache_read_input_token_cost_above_272k_tokens": 1e-06, - "cache_read_input_token_cost_priority": 1e-06, "cache_read_input_token_cost_above_272k_tokens_priority": 2e-06, + "cache_read_input_token_cost_priority": 1e-06, + "deprecation_date": "2027-10-26", "input_cost_per_token": 5e-06, "input_cost_per_token_above_272k_tokens": 1e-05, - "input_cost_per_token_priority": 1e-05, "input_cost_per_token_above_272k_tokens_priority": 2e-05, + "input_cost_per_token_priority": 1e-05, "litellm_provider": "azure_ai", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -3251,8 +3264,8 @@ "mode": "chat", "output_cost_per_token": 3e-05, "output_cost_per_token_above_272k_tokens": 4.5e-05, - "output_cost_per_token_priority": 6e-05, "output_cost_per_token_above_272k_tokens_priority": 9e-05, + "output_cost_per_token_priority": 6e-05, "source": "https://ai.azure.com/catalog/models/gpt-5.5", "supported_endpoints": [ "/v1/chat/completions", @@ -3267,7 +3280,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -3277,19 +3292,17 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure_ai/gpt-5.5-2026-04-23": { "cache_read_input_token_cost": 5e-07, "cache_read_input_token_cost_above_272k_tokens": 1e-06, - "cache_read_input_token_cost_priority": 1e-06, "cache_read_input_token_cost_above_272k_tokens_priority": 2e-06, + "cache_read_input_token_cost_priority": 1e-06, "input_cost_per_token": 5e-06, "input_cost_per_token_above_272k_tokens": 1e-05, - "input_cost_per_token_priority": 1e-05, "input_cost_per_token_above_272k_tokens_priority": 2e-05, + "input_cost_per_token_priority": 1e-05, "litellm_provider": "azure_ai", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -3297,8 +3310,8 @@ "mode": "chat", "output_cost_per_token": 3e-05, "output_cost_per_token_above_272k_tokens": 4.5e-05, - "output_cost_per_token_priority": 6e-05, "output_cost_per_token_above_272k_tokens_priority": 9e-05, + "output_cost_per_token_priority": 6e-05, "source": "https://ai.azure.com/catalog/models/gpt-5.5", "supported_endpoints": [ "/v1/chat/completions", @@ -3313,7 +3326,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -3323,20 +3338,18 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure_ai/gpt-5.4": { - "deprecation_date": "2027-09-02", "cache_read_input_token_cost": 2.5e-07, "cache_read_input_token_cost_above_272k_tokens": 5e-07, - "cache_read_input_token_cost_priority": 5e-07, "cache_read_input_token_cost_above_272k_tokens_priority": 1e-06, + "cache_read_input_token_cost_priority": 5e-07, + "deprecation_date": "2027-09-02", "input_cost_per_token": 2.5e-06, "input_cost_per_token_above_272k_tokens": 5e-06, - "input_cost_per_token_priority": 5e-06, "input_cost_per_token_above_272k_tokens_priority": 1e-05, + "input_cost_per_token_priority": 5e-06, "litellm_provider": "azure_ai", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -3344,8 +3357,8 @@ "mode": "chat", "output_cost_per_token": 1.5e-05, "output_cost_per_token_above_272k_tokens": 2.25e-05, - "output_cost_per_token_priority": 3e-05, "output_cost_per_token_above_272k_tokens_priority": 4.5e-05, + "output_cost_per_token_priority": 3e-05, "source": "https://ai.azure.com/catalog/models/gpt-5.4", "supported_endpoints": [ "/v1/chat/completions", @@ -3360,31 +3373,31 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true, - "supports_service_tier": true + "supports_xhigh_reasoning_effort": true }, "azure_ai/gpt-5.4-2026-03-05": { - "deprecation_date": "2027-09-02", "cache_read_input_token_cost": 2.5e-07, "cache_read_input_token_cost_above_272k_tokens": 5e-07, - "cache_read_input_token_cost_priority": 5e-07, "cache_read_input_token_cost_above_272k_tokens_priority": 1e-06, + "cache_read_input_token_cost_priority": 5e-07, + "deprecation_date": "2027-09-02", "input_cost_per_token": 2.5e-06, "input_cost_per_token_above_272k_tokens": 5e-06, - "input_cost_per_token_priority": 5e-06, "input_cost_per_token_above_272k_tokens_priority": 1e-05, + "input_cost_per_token_priority": 5e-06, "litellm_provider": "azure_ai", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -3392,8 +3405,8 @@ "mode": "chat", "output_cost_per_token": 1.5e-05, "output_cost_per_token_above_272k_tokens": 2.25e-05, - "output_cost_per_token_priority": 3e-05, "output_cost_per_token_above_272k_tokens_priority": 4.5e-05, + "output_cost_per_token_priority": 3e-05, "source": "https://ai.azure.com/catalog/models/gpt-5.4", "supported_endpoints": [ "/v1/chat/completions", @@ -3408,31 +3421,31 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true, - "supports_service_tier": true + "supports_xhigh_reasoning_effort": true }, "azure_ai/gpt-5.4-pro": { - "deprecation_date": "2027-09-07", "cache_read_input_token_cost": 3e-06, "cache_read_input_token_cost_above_272k_tokens": 6e-06, - "cache_read_input_token_cost_priority": 6e-06, "cache_read_input_token_cost_above_272k_tokens_priority": 1.2e-05, + "cache_read_input_token_cost_priority": 6e-06, + "deprecation_date": "2027-09-07", "input_cost_per_token": 3e-05, "input_cost_per_token_above_272k_tokens": 6e-05, - "input_cost_per_token_priority": 6e-05, "input_cost_per_token_above_272k_tokens_priority": 0.00012, + "input_cost_per_token_priority": 6e-05, "litellm_provider": "azure_ai", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -3440,8 +3453,8 @@ "mode": "responses", "output_cost_per_token": 0.00018, "output_cost_per_token_above_272k_tokens": 0.00027, - "output_cost_per_token_priority": 0.00036, "output_cost_per_token_above_272k_tokens_priority": 0.00054, + "output_cost_per_token_priority": 0.00036, "source": "https://ai.azure.com/catalog/models/gpt-5.4-pro", "supported_endpoints": [ "/v1/batch", @@ -3455,7 +3468,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -3465,20 +3480,18 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "azure_ai/gpt-5.4-pro-2026-03-05": { - "deprecation_date": "2027-09-07", "cache_read_input_token_cost": 3e-06, "cache_read_input_token_cost_above_272k_tokens": 6e-06, - "cache_read_input_token_cost_priority": 6e-06, "cache_read_input_token_cost_above_272k_tokens_priority": 1.2e-05, + "cache_read_input_token_cost_priority": 6e-06, + "deprecation_date": "2027-09-07", "input_cost_per_token": 3e-05, "input_cost_per_token_above_272k_tokens": 6e-05, - "input_cost_per_token_priority": 6e-05, "input_cost_per_token_above_272k_tokens_priority": 0.00012, + "input_cost_per_token_priority": 6e-05, "litellm_provider": "azure_ai", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -3486,8 +3499,8 @@ "mode": "responses", "output_cost_per_token": 0.00018, "output_cost_per_token_above_272k_tokens": 0.00027, - "output_cost_per_token_priority": 0.00036, "output_cost_per_token_above_272k_tokens_priority": 0.00054, + "output_cost_per_token_priority": 0.00036, "source": "https://ai.azure.com/catalog/models/gpt-5.4-pro", "supported_endpoints": [ "/v1/batch", @@ -3501,7 +3514,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -3511,14 +3526,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "azure_ai/gpt-5.4-mini": { - "deprecation_date": "2027-09-21", "cache_read_input_token_cost": 7.5e-08, "cache_read_input_token_cost_priority": 1.5e-07, + "deprecation_date": "2027-09-21", "input_cost_per_token": 7.5e-07, "input_cost_per_token_priority": 1.5e-06, "litellm_provider": "azure_ai", @@ -3542,25 +3555,25 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "supports_service_tier": true + "supports_xhigh_reasoning_effort": true }, "azure_ai/gpt-5.4-mini-2026-03-17": { - "deprecation_date": "2027-09-21", "cache_read_input_token_cost": 7.5e-08, "cache_read_input_token_cost_priority": 1.5e-07, + "deprecation_date": "2027-09-21", "input_cost_per_token": 7.5e-07, "input_cost_per_token_priority": 1.5e-06, "litellm_provider": "azure_ai", @@ -3584,25 +3597,25 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "supports_service_tier": true + "supports_xhigh_reasoning_effort": true }, "azure_ai/gpt-5.4-nano": { - "deprecation_date": "2027-09-21", "cache_read_input_token_cost": 2e-08, "cache_read_input_token_cost_priority": 4e-08, + "deprecation_date": "2027-09-21", "input_cost_per_token": 2e-07, "input_cost_per_token_priority": 4e-07, "litellm_provider": "azure_ai", @@ -3626,7 +3639,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -3636,14 +3651,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure_ai/gpt-5.4-nano-2026-03-17": { - "deprecation_date": "2027-09-21", "cache_read_input_token_cost": 2e-08, "cache_read_input_token_cost_priority": 4e-08, + "deprecation_date": "2027-09-21", "input_cost_per_token": 2e-07, "input_cost_per_token_priority": 4e-07, "litellm_provider": "azure_ai", @@ -3667,7 +3680,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -3677,21 +3692,19 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure_ai/model_router": { + "comment": "Flat cost of $0.14 per M input tokens for Azure AI Foundry Model Router infrastructure. Use pattern: azure_ai/model_router/ where deployment-name is your Azure deployment (e.g., azure-model-router)", "input_cost_per_token": 1.4e-07, - "output_cost_per_token": 0, "litellm_provider": "azure_ai", "mode": "chat", - "source": "https://azure.microsoft.com/en-us/pricing/details/ai-services/", - "comment": "Flat cost of $0.14 per M input tokens for Azure AI Foundry Model Router infrastructure. Use pattern: azure_ai/model_router/ where deployment-name is your Azure deployment (e.g., azure-model-router)" + "output_cost_per_token": 0, + "source": "https://azure.microsoft.com/en-us/pricing/details/ai-services/" }, "azure/eu/gpt-4o-2024-08-06": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 1.375e-06, + "deprecation_date": "2027-04-14", "input_cost_per_token": 2.75e-06, "litellm_provider": "azure", "max_input_tokens": 128000, @@ -3707,8 +3720,9 @@ "supports_vision": true }, "azure/eu/gpt-4o-2024-11-20": { - "deprecation_date": "2027-04-14", "cache_creation_input_token_cost": 1.38e-06, + "cache_read_input_token_cost": 1.375e-06, + "deprecation_date": "2027-04-14", "input_cost_per_token": 2.75e-06, "litellm_provider": "azure", "max_input_tokens": 128000, @@ -3720,8 +3734,7 @@ "supports_parallel_function_calling": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 1.375e-06 + "supports_vision": true }, "azure/eu/gpt-4o-mini-2024-07-18": { "cache_read_input_token_cost": 8.3e-08, @@ -3872,8 +3885,8 @@ "supports_vision": true }, "azure/eu/gpt-5.1": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 1.4e-07, + "deprecation_date": "2027-05-15", "input_cost_per_token": 1.38e-06, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -3896,6 +3909,7 @@ ], "supports_function_calling": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -3903,8 +3917,7 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_none_reasoning_effort": true + "supports_vision": true }, "azure/eu/gpt-5.1-chat": { "cache_read_input_token_cost": 1.4e-07, @@ -3931,6 +3944,7 @@ ], "supports_function_calling": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -3938,12 +3952,11 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_none_reasoning_effort": true + "supports_vision": true }, "azure/eu/gpt-5.1-codex": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 1.4e-07, + "deprecation_date": "2027-05-15", "input_cost_per_token": 1.38e-06, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -3973,8 +3986,8 @@ "supports_vision": true }, "azure/eu/gpt-5.1-codex-mini": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 2.8e-08, + "deprecation_date": "2027-05-15", "input_cost_per_token": 2.75e-07, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -4148,8 +4161,8 @@ "supports_vision": true }, "azure/global/gpt-4o-2024-08-06": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 1.25e-06, + "deprecation_date": "2027-04-14", "input_cost_per_token": 2.5e-06, "litellm_provider": "azure", "max_input_tokens": 128000, @@ -4165,8 +4178,8 @@ "supports_vision": true }, "azure/global/gpt-4o-2024-11-20": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 1.25e-06, + "deprecation_date": "2027-04-14", "input_cost_per_token": 2.5e-06, "litellm_provider": "azure", "max_input_tokens": 128000, @@ -4182,8 +4195,8 @@ "supports_vision": true }, "azure/global/gpt-5.1": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 1.25e-07, + "deprecation_date": "2027-05-15", "input_cost_per_token": 1.25e-06, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -4206,6 +4219,7 @@ ], "supports_function_calling": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -4213,8 +4227,7 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_none_reasoning_effort": true + "supports_vision": true }, "azure/global/gpt-5.1-chat": { "cache_read_input_token_cost": 1.25e-07, @@ -4241,6 +4254,7 @@ ], "supports_function_calling": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -4248,12 +4262,11 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_none_reasoning_effort": true + "supports_vision": true }, "azure/global/gpt-5.1-codex": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 1.25e-07, + "deprecation_date": "2027-05-15", "input_cost_per_token": 1.25e-06, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -4283,8 +4296,8 @@ "supports_vision": true }, "azure/global/gpt-5.1-codex-mini": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 2.5e-08, + "deprecation_date": "2027-05-15", "input_cost_per_token": 2.5e-07, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -4522,8 +4535,8 @@ "supports_vision": true }, "azure/gpt-4.1": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-04-14", "input_cost_per_token": 2e-06, "input_cost_per_token_batches": 1e-06, "litellm_provider": "azure", @@ -4556,8 +4569,8 @@ "supports_web_search": false }, "azure/gpt-4.1-2025-04-14": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-04-14", "input_cost_per_token": 2e-06, "input_cost_per_token_batches": 1e-06, "litellm_provider": "azure", @@ -4590,8 +4603,8 @@ "supports_web_search": false }, "azure/gpt-4.1-mini": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 1e-07, + "deprecation_date": "2027-04-14", "input_cost_per_token": 4e-07, "input_cost_per_token_batches": 2e-07, "litellm_provider": "azure", @@ -4624,8 +4637,8 @@ "supports_web_search": false }, "azure/gpt-4.1-mini-2025-04-14": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 1e-07, + "deprecation_date": "2027-04-14", "input_cost_per_token": 4e-07, "input_cost_per_token_batches": 2e-07, "litellm_provider": "azure", @@ -4658,8 +4671,8 @@ "supports_web_search": false }, "azure/gpt-4.1-nano": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 2.5e-08, + "deprecation_date": "2027-04-14", "input_cost_per_token": 1e-07, "input_cost_per_token_batches": 5e-08, "litellm_provider": "azure", @@ -4691,8 +4704,8 @@ "supports_vision": true }, "azure/gpt-4.1-nano-2025-04-14": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 2.5e-08, + "deprecation_date": "2027-04-14", "input_cost_per_token": 1e-07, "input_cost_per_token_batches": 5e-08, "litellm_provider": "azure", @@ -4774,8 +4787,8 @@ "supports_vision": true }, "azure/gpt-4o-2024-08-06": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 1.25e-06, + "deprecation_date": "2027-04-14", "input_cost_per_token": 2.5e-06, "litellm_provider": "azure", "max_input_tokens": 128000, @@ -4791,8 +4804,8 @@ "supports_vision": true }, "azure/gpt-4o-2024-11-20": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 1.25e-06, + "deprecation_date": "2027-04-14", "input_cost_per_token": 2.75e-06, "litellm_provider": "azure", "max_input_tokens": 128000, @@ -4861,6 +4874,8 @@ "text", "audio" ], + "supports_audio_input": true, + "supports_audio_output": true, "supports_function_calling": true, "supports_native_streaming": true, "supports_parallel_function_calling": true, @@ -4869,9 +4884,7 @@ "supports_response_schema": false, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": false, - "supports_audio_input": true, - "supports_audio_output": true + "supports_vision": false }, "azure/gpt-audio-mini": { "deprecation_date": "2027-04-06", @@ -4895,6 +4908,8 @@ "text", "audio" ], + "supports_audio_input": true, + "supports_audio_output": true, "supports_function_calling": true, "supports_native_streaming": true, "supports_parallel_function_calling": true, @@ -4903,9 +4918,7 @@ "supports_response_schema": false, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": false, - "supports_audio_input": true, - "supports_audio_output": true + "supports_vision": false }, "azure/gpt-audio-mini-2025-10-06": { "deprecation_date": "2027-04-06", @@ -4929,6 +4942,8 @@ "text", "audio" ], + "supports_audio_input": true, + "supports_audio_output": true, "supports_function_calling": true, "supports_native_streaming": true, "supports_parallel_function_calling": true, @@ -4937,9 +4952,7 @@ "supports_response_schema": false, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": false, - "supports_audio_input": true, - "supports_audio_output": true + "supports_vision": false }, "azure/gpt-4o-audio-preview-2024-12-17": { "input_cost_per_audio_token": 4e-05, @@ -4973,8 +4986,8 @@ "supports_vision": false }, "azure/gpt-4o-mini": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 7.5e-08, + "deprecation_date": "2027-04-14", "input_cost_per_token": 1.65e-07, "litellm_provider": "azure", "max_input_tokens": 128000, @@ -5061,6 +5074,7 @@ "cache_read_input_token_cost": 4e-07, "deprecation_date": "2027-03-02", "input_cost_per_audio_token": 3.2e-05, + "input_cost_per_image_token": 5e-06, "input_cost_per_token": 4e-06, "litellm_provider": "azure", "max_input_tokens": 32000, @@ -5086,14 +5100,14 @@ "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "input_cost_per_image_token": 5e-06 + "supports_tool_choice": true }, "azure/gpt-realtime-1.5-2026-02-23": { "cache_creation_input_audio_token_cost": 4e-07, "cache_read_input_token_cost": 4e-07, "deprecation_date": "2027-08-24", "input_cost_per_audio_token": 3.2e-05, + "input_cost_per_image_token": 5e-06, "input_cost_per_token": 4e-06, "litellm_provider": "azure", "max_input_tokens": 32000, @@ -5119,13 +5133,13 @@ "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "input_cost_per_image_token": 5e-06 + "supports_tool_choice": true }, "azure/gpt-realtime-mini": { "cache_creation_input_audio_token_cost": 3e-07, "cache_read_input_token_cost": 6e-08, "input_cost_per_audio_token": 1e-05, + "input_cost_per_image_token": 8e-07, "input_cost_per_token": 6e-07, "litellm_provider": "azure", "max_input_tokens": 32000, @@ -5151,13 +5165,13 @@ "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "input_cost_per_image_token": 8e-07 + "supports_tool_choice": true }, "azure/gpt-realtime-mini-2025-10-06": { "cache_creation_input_audio_token_cost": 3e-07, "cache_read_input_token_cost": 6e-08, "input_cost_per_audio_token": 1e-05, + "input_cost_per_image_token": 8e-07, "input_cost_per_token": 6e-07, "litellm_provider": "azure", "max_input_tokens": 32000, @@ -5183,8 +5197,7 @@ "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "input_cost_per_image_token": 8e-07 + "supports_tool_choice": true }, "azure/gpt-4o-mini-transcribe": { "input_cost_per_audio_token": 1.25e-06, @@ -5330,7 +5343,9 @@ "image" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -5338,9 +5353,7 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_none_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_vision": true }, "azure/gpt-5.1-chat-2025-11-13": { "cache_read_input_token_cost": 1.25e-07, @@ -5369,6 +5382,7 @@ ], "supports_function_calling": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": false, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -5376,8 +5390,7 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": false, - "supports_vision": true, - "supports_none_reasoning_effort": true + "supports_vision": true }, "azure/gpt-5.1-codex-2025-11-13": { "cache_read_input_token_cost": 1.25e-07, @@ -5448,8 +5461,8 @@ "supports_vision": true }, "azure/gpt-5": { - "deprecation_date": "2027-02-09", "cache_read_input_token_cost": 1.25e-07, + "deprecation_date": "2027-02-09", "input_cost_per_token": 1.25e-06, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -5612,8 +5625,8 @@ "supports_vision": true }, "azure/gpt-5-mini": { - "deprecation_date": "2027-02-09", "cache_read_input_token_cost": 2.5e-08, + "deprecation_date": "2027-02-09", "input_cost_per_token": 2.5e-07, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -5678,8 +5691,8 @@ "supports_vision": true }, "azure/gpt-5-nano": { - "deprecation_date": "2027-02-09", "cache_read_input_token_cost": 5e-09, + "deprecation_date": "2027-02-09", "input_cost_per_token": 5e-08, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -5774,8 +5787,8 @@ "supports_vision": true }, "azure/gpt-5.1": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 1.25e-07, + "deprecation_date": "2027-05-15", "input_cost_per_token": 1.25e-06, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -5798,6 +5811,7 @@ ], "supports_function_calling": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -5805,8 +5819,7 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_none_reasoning_effort": true + "supports_vision": true }, "azure/gpt-5.1-chat": { "cache_read_input_token_cost": 1.25e-07, @@ -5833,6 +5846,7 @@ ], "supports_function_calling": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -5840,12 +5854,11 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_none_reasoning_effort": true + "supports_vision": true }, "azure/gpt-5.1-codex": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 1.25e-07, + "deprecation_date": "2027-05-15", "input_cost_per_token": 1.25e-06, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -5906,8 +5919,8 @@ "supports_vision": true }, "azure/gpt-5.1-codex-mini": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 2.5e-08, + "deprecation_date": "2027-05-15", "input_cost_per_token": 2.5e-07, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -5937,8 +5950,8 @@ "supports_vision": true }, "azure/gpt-5.2": { - "deprecation_date": "2027-06-08", "cache_read_input_token_cost": 1.75e-07, + "deprecation_date": "2027-06-08", "input_cost_per_token": 1.75e-06, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -6247,15 +6260,15 @@ "supports_web_search": true }, "azure/gpt-5.4": { - "deprecation_date": "2027-09-02", "cache_read_input_token_cost": 2.5e-07, "cache_read_input_token_cost_above_272k_tokens": 5e-07, - "cache_read_input_token_cost_priority": 5e-07, "cache_read_input_token_cost_above_272k_tokens_priority": 1e-06, + "cache_read_input_token_cost_priority": 5e-07, + "deprecation_date": "2027-09-02", "input_cost_per_token": 2.5e-06, "input_cost_per_token_above_272k_tokens": 5e-06, - "input_cost_per_token_priority": 5e-06, "input_cost_per_token_above_272k_tokens_priority": 1e-05, + "input_cost_per_token_priority": 5e-06, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -6263,8 +6276,8 @@ "mode": "chat", "output_cost_per_token": 1.5e-05, "output_cost_per_token_above_272k_tokens": 2.25e-05, - "output_cost_per_token_priority": 3e-05, "output_cost_per_token_above_272k_tokens_priority": 4.5e-05, + "output_cost_per_token_priority": 3e-05, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6278,7 +6291,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6287,23 +6302,21 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "azure/us/gpt-5.4": { - "deprecation_date": "2027-09-02", "cache_read_input_token_cost": 2.8e-07, "cache_read_input_token_cost_priority": 5.5e-07, + "deprecation_date": "2027-09-02", "input_cost_per_token": 2.75e-06, "input_cost_per_token_priority": 5.5e-06, - "output_cost_per_token": 1.65e-05, - "output_cost_per_token_priority": 3.3e-05, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1.65e-05, + "output_cost_per_token_priority": 3.3e-05, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6317,7 +6330,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6326,23 +6341,21 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "azure/eu/gpt-5.4": { - "deprecation_date": "2027-09-02", "cache_read_input_token_cost": 2.8e-07, "cache_read_input_token_cost_priority": 5.5e-07, + "deprecation_date": "2027-09-02", "input_cost_per_token": 2.75e-06, "input_cost_per_token_priority": 5.5e-06, - "output_cost_per_token": 1.65e-05, - "output_cost_per_token_priority": 3.3e-05, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1.65e-05, + "output_cost_per_token_priority": 3.3e-05, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6356,7 +6369,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6365,20 +6380,18 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "azure/gpt-5.4-2026-03-05": { "cache_read_input_token_cost": 2.5e-07, "cache_read_input_token_cost_above_272k_tokens": 5e-07, - "cache_read_input_token_cost_priority": 5e-07, "cache_read_input_token_cost_above_272k_tokens_priority": 1e-06, + "cache_read_input_token_cost_priority": 5e-07, "deprecation_date": "2027-09-02", "input_cost_per_token": 2.5e-06, "input_cost_per_token_above_272k_tokens": 5e-06, - "input_cost_per_token_priority": 5e-06, "input_cost_per_token_above_272k_tokens_priority": 1e-05, + "input_cost_per_token_priority": 5e-06, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -6386,8 +6399,8 @@ "mode": "chat", "output_cost_per_token": 1.5e-05, "output_cost_per_token_above_272k_tokens": 2.25e-05, - "output_cost_per_token_priority": 3e-05, "output_cost_per_token_above_272k_tokens_priority": 4.5e-05, + "output_cost_per_token_priority": 3e-05, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6401,7 +6414,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6410,9 +6425,7 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "azure/us/gpt-5.4-2026-03-05": { "cache_read_input_token_cost": 2.8e-07, @@ -6420,13 +6433,13 @@ "deprecation_date": "2027-09-02", "input_cost_per_token": 2.75e-06, "input_cost_per_token_priority": 5.5e-06, - "output_cost_per_token": 1.65e-05, - "output_cost_per_token_priority": 3.3e-05, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1.65e-05, + "output_cost_per_token_priority": 3.3e-05, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6440,7 +6453,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6449,9 +6464,7 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "azure/eu/gpt-5.4-2026-03-05": { "cache_read_input_token_cost": 2.8e-07, @@ -6459,13 +6472,13 @@ "deprecation_date": "2027-09-02", "input_cost_per_token": 2.75e-06, "input_cost_per_token_priority": 5.5e-06, - "output_cost_per_token": 1.65e-05, - "output_cost_per_token_priority": 3.3e-05, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1.65e-05, + "output_cost_per_token_priority": 3.3e-05, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6479,7 +6492,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6488,14 +6503,12 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "azure/gpt-5.4-pro": { - "deprecation_date": "2027-09-07", "cache_read_input_token_cost": 3e-06, "cache_read_input_token_cost_above_272k_tokens": 6e-06, + "deprecation_date": "2027-09-07", "input_cost_per_token": 3e-05, "input_cost_per_token_above_272k_tokens": 6e-05, "litellm_provider": "azure", @@ -6575,6 +6588,9 @@ "supports_web_search": true }, "azure/gpt-5.6": { + "cache_creation_input_token_cost": 6.25e-06, + "cache_creation_input_token_cost_above_272k_tokens": 1.25e-05, + "cache_creation_input_token_cost_priority": 1.25e-05, "cache_read_input_token_cost": 5e-07, "cache_read_input_token_cost_above_272k_tokens": 1e-06, "cache_read_input_token_cost_priority": 1e-06, @@ -6590,6 +6606,7 @@ "output_cost_per_token_above_272k_tokens": 4.5e-05, "output_cost_per_token_priority": 6e-05, "search_context_cost_per_query": {}, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6603,7 +6620,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6613,15 +6632,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, + "supports_xhigh_reasoning_effort": true + }, + "azure/gpt-5.6-sol": { "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_272k_tokens": 1.25e-05, "cache_creation_input_token_cost_priority": 1.25e-05, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service" - }, - "azure/gpt-5.6-sol": { "cache_read_input_token_cost": 5e-07, "cache_read_input_token_cost_above_272k_tokens": 1e-06, "cache_read_input_token_cost_priority": 1e-06, @@ -6642,6 +6658,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6655,7 +6672,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6665,15 +6684,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "cache_creation_input_token_cost": 6.25e-06, - "cache_creation_input_token_cost_above_272k_tokens": 1.25e-05, - "cache_creation_input_token_cost_priority": 1.25e-05, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service" + "supports_xhigh_reasoning_effort": true }, "azure/gpt-5.6-terra": { + "cache_creation_input_token_cost": 2.5e-06, + "cache_creation_input_token_cost_above_272k_tokens": 5e-06, + "cache_creation_input_token_cost_priority": 5e-06, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_272k_tokens": 4e-07, "cache_read_input_token_cost_priority": 4e-07, @@ -6694,6 +6710,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6707,7 +6724,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6717,15 +6736,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "cache_creation_input_token_cost": 2.5e-06, - "cache_creation_input_token_cost_above_272k_tokens": 5e-06, - "cache_creation_input_token_cost_priority": 5e-06, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service" + "supports_xhigh_reasoning_effort": true }, "azure/gpt-5.6-luna": { + "cache_creation_input_token_cost": 2.5e-07, + "cache_creation_input_token_cost_above_272k_tokens": 5e-07, + "cache_creation_input_token_cost_priority": 5e-07, "cache_read_input_token_cost": 2e-08, "cache_read_input_token_cost_above_272k_tokens": 4e-08, "cache_read_input_token_cost_priority": 4e-08, @@ -6746,6 +6762,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6759,7 +6776,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6769,15 +6788,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "cache_creation_input_token_cost": 2.5e-07, - "cache_creation_input_token_cost_above_272k_tokens": 5e-07, - "cache_creation_input_token_cost_priority": 5e-07, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service" + "supports_xhigh_reasoning_effort": true }, "azure/us/gpt-5.6": { + "cache_creation_input_token_cost": 6.875e-06, + "cache_creation_input_token_cost_above_272k_tokens": 1.375e-05, + "cache_creation_input_token_cost_priority": 1.375e-05, "cache_read_input_token_cost": 5.5e-07, "cache_read_input_token_cost_above_272k_tokens": 1.1e-06, "cache_read_input_token_cost_priority": 1.1e-06, @@ -6797,6 +6813,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6810,7 +6827,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6820,15 +6839,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "cache_creation_input_token_cost": 6.875e-06, - "cache_creation_input_token_cost_above_272k_tokens": 1.375e-05, - "cache_creation_input_token_cost_priority": 1.375e-05, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service" + "supports_xhigh_reasoning_effort": true }, "azure/us/gpt-5.6-sol": { + "cache_creation_input_token_cost": 6.875e-06, + "cache_creation_input_token_cost_above_272k_tokens": 1.375e-05, + "cache_creation_input_token_cost_priority": 1.375e-05, "cache_read_input_token_cost": 5.5e-07, "cache_read_input_token_cost_above_272k_tokens": 1.1e-06, "cache_read_input_token_cost_priority": 1.1e-06, @@ -6849,6 +6865,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6862,7 +6879,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6872,15 +6891,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "cache_creation_input_token_cost": 6.875e-06, - "cache_creation_input_token_cost_above_272k_tokens": 1.375e-05, - "cache_creation_input_token_cost_priority": 1.375e-05, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service" + "supports_xhigh_reasoning_effort": true }, "azure/us/gpt-5.6-terra": { + "cache_creation_input_token_cost": 2.75e-06, + "cache_creation_input_token_cost_above_272k_tokens": 5.5e-06, + "cache_creation_input_token_cost_priority": 5.5e-06, "cache_read_input_token_cost": 2.2e-07, "cache_read_input_token_cost_above_272k_tokens": 4.4e-07, "cache_read_input_token_cost_priority": 4.4e-07, @@ -6901,6 +6917,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6914,7 +6931,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6924,15 +6943,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "cache_creation_input_token_cost_above_272k_tokens": 5.5e-06, - "cache_creation_input_token_cost": 2.75e-06, - "cache_creation_input_token_cost_priority": 5.5e-06, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service" + "supports_xhigh_reasoning_effort": true }, "azure/us/gpt-5.6-luna": { + "cache_creation_input_token_cost": 2.75e-07, + "cache_creation_input_token_cost_above_272k_tokens": 5.5e-07, + "cache_creation_input_token_cost_priority": 5.5e-07, "cache_read_input_token_cost": 2.2e-08, "cache_read_input_token_cost_above_272k_tokens": 4.4e-08, "cache_read_input_token_cost_priority": 4.4e-08, @@ -6953,6 +6969,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -6966,7 +6983,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -6976,15 +6995,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "cache_creation_input_token_cost": 2.75e-07, - "cache_creation_input_token_cost_above_272k_tokens": 5.5e-07, - "cache_creation_input_token_cost_priority": 5.5e-07, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service" + "supports_xhigh_reasoning_effort": true }, "azure/eu/gpt-5.6": { + "cache_creation_input_token_cost": 6.875e-06, + "cache_creation_input_token_cost_above_272k_tokens": 1.375e-05, + "cache_creation_input_token_cost_priority": 1.375e-05, "cache_read_input_token_cost": 5.5e-07, "cache_read_input_token_cost_above_272k_tokens": 1.1e-06, "cache_read_input_token_cost_priority": 1.1e-06, @@ -7004,6 +7020,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -7017,7 +7034,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7027,15 +7046,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, + "supports_xhigh_reasoning_effort": true + }, + "azure/eu/gpt-5.6-sol": { "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_272k_tokens": 1.375e-05, "cache_creation_input_token_cost_priority": 1.375e-05, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service" - }, - "azure/eu/gpt-5.6-sol": { "cache_read_input_token_cost": 5.5e-07, "cache_read_input_token_cost_above_272k_tokens": 1.1e-06, "cache_read_input_token_cost_priority": 1.1e-06, @@ -7056,6 +7072,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -7069,7 +7086,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7079,15 +7098,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "cache_creation_input_token_cost": 6.875e-06, - "cache_creation_input_token_cost_above_272k_tokens": 1.375e-05, - "cache_creation_input_token_cost_priority": 1.375e-05, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service" + "supports_xhigh_reasoning_effort": true }, "azure/eu/gpt-5.6-terra": { + "cache_creation_input_token_cost": 2.75e-06, + "cache_creation_input_token_cost_above_272k_tokens": 5.5e-06, + "cache_creation_input_token_cost_priority": 5.5e-06, "cache_read_input_token_cost": 2.2e-07, "cache_read_input_token_cost_above_272k_tokens": 4.4e-07, "cache_read_input_token_cost_priority": 4.4e-07, @@ -7108,6 +7124,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -7121,7 +7138,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7131,15 +7150,12 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "cache_creation_input_token_cost_above_272k_tokens": 5.5e-06, - "cache_creation_input_token_cost_priority": 5.5e-06, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", - "cache_creation_input_token_cost": 2.75e-06 + "supports_xhigh_reasoning_effort": true }, "azure/eu/gpt-5.6-luna": { + "cache_creation_input_token_cost": 2.75e-07, + "cache_creation_input_token_cost_above_272k_tokens": 5.5e-07, + "cache_creation_input_token_cost_priority": 5.5e-07, "cache_read_input_token_cost": 2.2e-08, "cache_read_input_token_cost_above_272k_tokens": 4.4e-08, "cache_read_input_token_cost_priority": 4.4e-08, @@ -7160,6 +7176,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -7173,7 +7190,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7183,24 +7202,18 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "cache_creation_input_token_cost": 2.75e-07, - "cache_creation_input_token_cost_above_272k_tokens": 5.5e-07, - "cache_creation_input_token_cost_priority": 5.5e-07, - "source": "https://azure.microsoft.com/en-us/products/ai-services/openai-service" + "supports_xhigh_reasoning_effort": true }, "azure/gpt-5.5": { - "deprecation_date": "2027-10-26", "cache_read_input_token_cost": 5e-07, "cache_read_input_token_cost_above_272k_tokens": 1e-06, - "cache_read_input_token_cost_priority": 1e-06, "cache_read_input_token_cost_above_272k_tokens_priority": 2e-06, + "cache_read_input_token_cost_priority": 1e-06, + "deprecation_date": "2027-10-26", "input_cost_per_token": 5e-06, "input_cost_per_token_above_272k_tokens": 1e-05, - "input_cost_per_token_priority": 1e-05, "input_cost_per_token_above_272k_tokens_priority": 2e-05, + "input_cost_per_token_priority": 1e-05, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -7208,8 +7221,8 @@ "mode": "chat", "output_cost_per_token": 3e-05, "output_cost_per_token_above_272k_tokens": 4.5e-05, - "output_cost_per_token_priority": 6e-05, "output_cost_per_token_above_272k_tokens_priority": 9e-05, + "output_cost_per_token_priority": 6e-05, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -7228,7 +7241,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7238,26 +7253,24 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure/us/gpt-5.5": { - "deprecation_date": "2027-10-26", "cache_read_input_token_cost": 5.5e-07, "cache_read_input_token_cost_above_272k_tokens": 1.1e-06, "cache_read_input_token_cost_priority": 1.38e-06, + "deprecation_date": "2027-10-26", "input_cost_per_token": 5.5e-06, "input_cost_per_token_above_272k_tokens": 1.1e-05, "input_cost_per_token_priority": 1.375e-05, - "output_cost_per_token": 3.3e-05, - "output_cost_per_token_above_272k_tokens": 4.95e-05, - "output_cost_per_token_priority": 8.25e-05, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.3e-05, + "output_cost_per_token_above_272k_tokens": 4.95e-05, + "output_cost_per_token_priority": 8.25e-05, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -7276,7 +7289,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7286,26 +7301,24 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure/eu/gpt-5.5": { - "deprecation_date": "2027-10-26", "cache_read_input_token_cost": 5.5e-07, "cache_read_input_token_cost_above_272k_tokens": 1.1e-06, "cache_read_input_token_cost_priority": 1.38e-06, + "deprecation_date": "2027-10-26", "input_cost_per_token": 5.5e-06, "input_cost_per_token_above_272k_tokens": 1.1e-05, "input_cost_per_token_priority": 1.375e-05, - "output_cost_per_token": 3.3e-05, - "output_cost_per_token_above_272k_tokens": 4.95e-05, - "output_cost_per_token_priority": 8.25e-05, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.3e-05, + "output_cost_per_token_above_272k_tokens": 4.95e-05, + "output_cost_per_token_priority": 8.25e-05, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -7324,7 +7337,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7334,19 +7349,17 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure/gpt-5.5-2026-04-23": { "cache_read_input_token_cost": 5e-07, "cache_read_input_token_cost_above_272k_tokens": 1e-06, - "cache_read_input_token_cost_priority": 1e-06, "cache_read_input_token_cost_above_272k_tokens_priority": 2e-06, + "cache_read_input_token_cost_priority": 1e-06, "input_cost_per_token": 5e-06, "input_cost_per_token_above_272k_tokens": 1e-05, - "input_cost_per_token_priority": 1e-05, "input_cost_per_token_above_272k_tokens_priority": 2e-05, + "input_cost_per_token_priority": 1e-05, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -7354,8 +7367,8 @@ "mode": "chat", "output_cost_per_token": 3e-05, "output_cost_per_token_above_272k_tokens": 4.5e-05, - "output_cost_per_token_priority": 6e-05, "output_cost_per_token_above_272k_tokens_priority": 9e-05, + "output_cost_per_token_priority": 6e-05, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -7392,14 +7405,14 @@ "input_cost_per_token": 5.5e-06, "input_cost_per_token_above_272k_tokens": 1.1e-05, "input_cost_per_token_priority": 1.375e-05, - "output_cost_per_token": 3.3e-05, - "output_cost_per_token_above_272k_tokens": 4.95e-05, - "output_cost_per_token_priority": 8.25e-05, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.3e-05, + "output_cost_per_token_above_272k_tokens": 4.95e-05, + "output_cost_per_token_priority": 8.25e-05, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -7436,14 +7449,14 @@ "input_cost_per_token": 5.5e-06, "input_cost_per_token_above_272k_tokens": 1.1e-05, "input_cost_per_token_priority": 1.375e-05, - "output_cost_per_token": 3.3e-05, - "output_cost_per_token_above_272k_tokens": 4.95e-05, - "output_cost_per_token_priority": 8.25e-05, "litellm_provider": "azure", "max_input_tokens": 1050000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.3e-05, + "output_cost_per_token_above_272k_tokens": 4.95e-05, + "output_cost_per_token_priority": 8.25e-05, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -7502,7 +7515,10 @@ "text" ], "supports_function_calling": true, + "supports_low_reasoning_effort": false, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7512,10 +7528,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "supports_low_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure/gpt-5.5-pro-2026-04-23": { "cache_read_input_token_cost": 3e-06, @@ -7558,8 +7571,8 @@ "supports_web_search": true }, "azure/gpt-5.4-mini": { - "deprecation_date": "2027-09-21", "cache_read_input_token_cost": 7.5e-08, + "deprecation_date": "2027-09-21", "input_cost_per_token": 7.5e-07, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -7585,7 +7598,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7595,9 +7610,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure/gpt-5.4-mini-2026-03-17": { "cache_read_input_token_cost": 7.5e-08, @@ -7627,7 +7640,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7637,13 +7652,11 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure/gpt-5.4-nano": { - "deprecation_date": "2027-09-21", "cache_read_input_token_cost": 2e-08, + "deprecation_date": "2027-09-21", "input_cost_per_token": 2e-07, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -7669,7 +7682,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7679,9 +7694,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure/gpt-5.4-nano-2026-03-17": { "cache_read_input_token_cost": 2e-08, @@ -7711,7 +7724,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -7721,9 +7736,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "azure/gpt-image-1": { "cache_read_input_token_cost": 1.25e-06, @@ -7860,64 +7873,64 @@ ] }, "azure/gpt-image-1.5": { - "deprecation_date": "2027-06-16", "cache_read_input_token_cost": 1.25e-06, - "input_cost_per_token": 5e-06, + "deprecation_date": "2027-06-16", "input_cost_per_image_token": 8e-06, + "input_cost_per_token": 5e-06, "litellm_provider": "azure", "mode": "image_generation", "output_cost_per_image_token": 3.2e-05, + "output_cost_per_token": 1e-05, "supported_endpoints": [ "/v1/images/generations", "/v1/images/edits" - ], - "output_cost_per_token": 1e-05 + ] }, "azure/gpt-image-1.5-2025-12-16": { "cache_read_input_token_cost": 1.25e-06, "deprecation_date": "2027-06-16", - "input_cost_per_token": 5e-06, "input_cost_per_image_token": 8e-06, + "input_cost_per_token": 5e-06, "litellm_provider": "azure", "mode": "image_generation", "output_cost_per_image_token": 3.2e-05, + "output_cost_per_token": 1e-05, "supported_endpoints": [ "/v1/images/generations", "/v1/images/edits" - ], - "output_cost_per_token": 1e-05 + ] }, "azure/gpt-image-2": { - "deprecation_date": "2027-10-21", "cache_read_input_token_cost": 1.25e-06, - "input_cost_per_token": 5e-06, + "deprecation_date": "2027-10-21", "input_cost_per_image_token": 8e-06, + "input_cost_per_token": 5e-06, "litellm_provider": "azure", "mode": "image_generation", - "output_cost_per_token": 1e-05, "output_cost_per_image_token": 3e-05, + "output_cost_per_token": 1e-05, "supported_endpoints": [ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "azure/gpt-image-2-2026-04-21": { "cache_read_input_token_cost": 1.25e-06, "deprecation_date": "2027-10-21", - "input_cost_per_token": 5e-06, "input_cost_per_image_token": 8e-06, + "input_cost_per_token": 5e-06, "litellm_provider": "azure", "mode": "image_generation", - "output_cost_per_token": 1e-05, "output_cost_per_image_token": 3e-05, + "output_cost_per_token": 1e-05, "supported_endpoints": [ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "azure/low/1024-x-1024/gpt-image-1-mini": { "deprecation_date": "2027-04-07", @@ -8028,8 +8041,8 @@ "supports_function_calling": true }, "azure/o1": { - "deprecation_date": "2026-10-21", "cache_read_input_token_cost": 7.5e-06, + "deprecation_date": "2026-10-21", "input_cost_per_token": 1.5e-05, "litellm_provider": "azure", "max_input_tokens": 200000, @@ -8123,8 +8136,8 @@ "supports_vision": false }, "azure/o3": { - "deprecation_date": "2026-10-21", "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2026-10-21", "input_cost_per_token": 2e-06, "litellm_provider": "azure", "max_input_tokens": 200000, @@ -8153,8 +8166,8 @@ "supports_vision": true }, "azure/o3-2025-04-16": { - "deprecation_date": "2026-10-21", "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2026-10-21", "input_cost_per_token": 2e-06, "litellm_provider": "azure", "max_input_tokens": 200000, @@ -8221,8 +8234,8 @@ "supports_web_search": true }, "azure/o3-mini": { - "deprecation_date": "2026-10-01", "cache_read_input_token_cost": 5.5e-07, + "deprecation_date": "2026-10-01", "input_cost_per_token": 1.1e-06, "litellm_provider": "azure", "max_input_tokens": 200000, @@ -8314,8 +8327,8 @@ "supports_vision": true }, "azure/o4-mini": { - "deprecation_date": "2026-10-16", "cache_read_input_token_cost": 2.75e-07, + "deprecation_date": "2026-10-16", "input_cost_per_token": 1.1e-06, "litellm_provider": "azure", "max_input_tokens": 200000, @@ -8448,8 +8461,8 @@ "mode": "audio_speech" }, "azure/us/gpt-4.1-2025-04-14": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 5.5e-07, + "deprecation_date": "2027-04-14", "input_cost_per_token": 2.2e-06, "input_cost_per_token_batches": 1.1e-06, "litellm_provider": "azure", @@ -8482,8 +8495,8 @@ "supports_web_search": false }, "azure/us/gpt-4.1-mini-2025-04-14": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 1.1e-07, + "deprecation_date": "2027-04-14", "input_cost_per_token": 4.4e-07, "input_cost_per_token_batches": 2.2e-07, "litellm_provider": "azure", @@ -8516,8 +8529,8 @@ "supports_web_search": false }, "azure/us/gpt-4.1-nano-2025-04-14": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 2.5e-08, + "deprecation_date": "2027-04-14", "input_cost_per_token": 1.1e-07, "input_cost_per_token_batches": 6e-08, "litellm_provider": "azure", @@ -8549,8 +8562,8 @@ "supports_vision": true }, "azure/us/gpt-4o-2024-08-06": { - "deprecation_date": "2027-04-14", "cache_read_input_token_cost": 1.375e-06, + "deprecation_date": "2027-04-14", "input_cost_per_token": 2.75e-06, "litellm_provider": "azure", "max_input_tokens": 128000, @@ -8566,8 +8579,9 @@ "supports_vision": true }, "azure/us/gpt-4o-2024-11-20": { - "deprecation_date": "2027-04-14", "cache_creation_input_token_cost": 1.38e-06, + "cache_read_input_token_cost": 1.375e-06, + "deprecation_date": "2027-04-14", "input_cost_per_token": 2.75e-06, "litellm_provider": "azure", "max_input_tokens": 128000, @@ -8579,8 +8593,7 @@ "supports_parallel_function_calling": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 1.375e-06 + "supports_vision": true }, "azure/us/gpt-4o-mini-2024-07-18": { "cache_read_input_token_cost": 8.3e-08, @@ -8764,8 +8777,8 @@ "supports_vision": true }, "azure/us/gpt-5.1": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 1.4e-07, + "deprecation_date": "2027-05-15", "input_cost_per_token": 1.38e-06, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -8788,6 +8801,7 @@ ], "supports_function_calling": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -8795,8 +8809,7 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_none_reasoning_effort": true + "supports_vision": true }, "azure/us/gpt-5.1-chat": { "cache_read_input_token_cost": 1.4e-07, @@ -8823,6 +8836,7 @@ ], "supports_function_calling": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -8830,12 +8844,11 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_none_reasoning_effort": true + "supports_vision": true }, "azure/us/gpt-5.1-codex": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 1.4e-07, + "deprecation_date": "2027-05-15", "input_cost_per_token": 1.38e-06, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -8865,8 +8878,8 @@ "supports_vision": true }, "azure/us/gpt-5.1-codex-mini": { - "deprecation_date": "2027-05-15", "cache_read_input_token_cost": 2.8e-08, + "deprecation_date": "2027-05-15", "input_cost_per_token": 2.75e-07, "litellm_provider": "azure", "max_input_tokens": 272000, @@ -8942,8 +8955,8 @@ "supports_vision": false }, "azure/us/o3-2025-04-16": { - "deprecation_date": "2026-10-21", "cache_read_input_token_cost": 5.5e-07, + "deprecation_date": "2026-10-21", "input_cost_per_token": 2.2e-06, "litellm_provider": "azure", "max_input_tokens": 200000, @@ -9063,8 +9076,8 @@ ] }, "azure_ai/FW-DeepSeek-V3.2": { - "deprecation_date": "2027-07-01", "cache_read_input_token_cost": 3.1e-07, + "deprecation_date": "2027-07-01", "input_cost_per_token": 6.2e-07, "litellm_provider": "azure_ai", "max_input_tokens": 163840, @@ -9094,8 +9107,8 @@ "supports_tool_choice": true }, "azure_ai/FW-GLM-5": { - "deprecation_date": "2027-07-01", "cache_read_input_token_cost": 2.2e-07, + "deprecation_date": "2027-07-01", "input_cost_per_token": 1.1e-06, "litellm_provider": "azure_ai", "max_input_tokens": 200000, @@ -9110,8 +9123,8 @@ "supports_tool_choice": true }, "azure_ai/FW-GLM-5.1": { - "deprecation_date": "2027-07-01", "cache_read_input_token_cost": 2.86e-07, + "deprecation_date": "2027-07-01", "input_cost_per_token": 1.54e-06, "litellm_provider": "azure_ai", "max_input_tokens": 202800, @@ -9177,8 +9190,8 @@ "supports_tool_choice": true }, "azure_ai/FW-Kimi-K2.5": { - "deprecation_date": "2027-07-01", "cache_read_input_token_cost": 1.1e-07, + "deprecation_date": "2027-07-01", "input_cost_per_token": 6.6e-07, "litellm_provider": "azure_ai", "max_input_tokens": 262144, @@ -9270,8 +9283,8 @@ "supports_vision": true }, "azure_ai/FW-MiniMax-M2.5": { - "deprecation_date": "2027-07-01", "cache_read_input_token_cost": 3.3e-08, + "deprecation_date": "2027-07-01", "input_cost_per_token": 3.3e-07, "litellm_provider": "azure_ai", "max_input_tokens": 1000000, @@ -9645,54 +9658,54 @@ "output_cost_per_token": 5e-07, "source": "https://azure.microsoft.com/en-us/pricing/details/ai-foundry-models/microsoft/", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_reasoning": true + "supports_reasoning": true, + "supports_tool_choice": true }, "azure_ai/mistral-document-ai-2505": { "deprecation_date": "2026-07-20", "litellm_provider": "azure_ai", - "ocr_cost_per_page": 0.003, "mode": "ocr", + "ocr_cost_per_page": 0.003, + "source": "https://devblogs.microsoft.com/foundry/whats-new-in-azure-ai-foundry-august-2025/#mistral-document-ai-(ocr)-%E2%80%94-serverless-in-foundry", "supported_endpoints": [ "/v1/ocr" - ], - "source": "https://devblogs.microsoft.com/foundry/whats-new-in-azure-ai-foundry-august-2025/#mistral-document-ai-(ocr)-%E2%80%94-serverless-in-foundry" + ] }, "azure_ai/mistral-document-ai-2512": { "litellm_provider": "azure_ai", - "ocr_cost_per_page": 0.003, "mode": "ocr", + "ocr_cost_per_page": 0.003, + "source": "https://azure.microsoft.com/en-us/pricing/details/ai-foundry-models/", "supported_endpoints": [ "/v1/ocr" - ], - "source": "https://azure.microsoft.com/en-us/pricing/details/ai-foundry-models/" + ] }, "azure_ai/doc-intelligence/prebuilt-read": { "litellm_provider": "azure_ai", - "ocr_cost_per_page": 0.0015, "mode": "ocr", + "ocr_cost_per_page": 0.0015, + "source": "https://azure.microsoft.com/en-us/pricing/details/ai-document-intelligence/", "supported_endpoints": [ "/v1/ocr" - ], - "source": "https://azure.microsoft.com/en-us/pricing/details/ai-document-intelligence/" + ] }, "azure_ai/doc-intelligence/prebuilt-layout": { "litellm_provider": "azure_ai", - "ocr_cost_per_page": 0.01, "mode": "ocr", + "ocr_cost_per_page": 0.01, + "source": "https://azure.microsoft.com/en-us/pricing/details/ai-document-intelligence/", "supported_endpoints": [ "/v1/ocr" - ], - "source": "https://azure.microsoft.com/en-us/pricing/details/ai-document-intelligence/" + ] }, "azure_ai/doc-intelligence/prebuilt-document": { "litellm_provider": "azure_ai", - "ocr_cost_per_page": 0.01, "mode": "ocr", + "ocr_cost_per_page": 0.01, + "source": "https://azure.microsoft.com/en-us/pricing/details/ai-document-intelligence/", "supported_endpoints": [ "/v1/ocr" - ], - "source": "https://azure.microsoft.com/en-us/pricing/details/ai-document-intelligence/" + ] }, "azure_ai/MAI-DS-R1": { "input_cost_per_token": 1.35e-06, @@ -9983,12 +9996,12 @@ "azure_ai/grok-4-fast-non-reasoning": { "deprecation_date": "2026-05-01", "input_cost_per_token": 2e-07, - "output_cost_per_token": 5e-07, "litellm_provider": "azure_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 5e-07, "supports_function_calling": true, "supports_response_schema": true, "supports_tool_choice": true, @@ -9997,12 +10010,12 @@ "azure_ai/grok-4-fast-reasoning": { "deprecation_date": "2026-05-01", "input_cost_per_token": 2e-07, - "output_cost_per_token": 5e-07, "litellm_provider": "azure_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 5e-07, "source": "https://azure.microsoft.com/en-us/pricing/details/ai-foundry-models/grok/", "supports_function_calling": true, "supports_response_schema": true, @@ -10011,12 +10024,12 @@ }, "azure_ai/grok-4-1-fast-non-reasoning": { "input_cost_per_token": 2e-07, - "output_cost_per_token": 5e-07, "litellm_provider": "azure_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 5e-07, "source": "https://techcommunity.microsoft.com/t5/Azure-AI-Foundry-Blog/Grok-4-0-Goes-GA-in-Microsoft-Foundry-and-Grok-4-1-Fast-Arrives/ba-p/4497964", "supports_function_calling": true, "supports_response_schema": true, @@ -10025,12 +10038,12 @@ }, "azure_ai/grok-4-1-fast-reasoning": { "input_cost_per_token": 2e-07, - "output_cost_per_token": 5e-07, "litellm_provider": "azure_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 5e-07, "source": "https://techcommunity.microsoft.com/t5/Azure-AI-Foundry-Blog/Grok-4-0-Goes-GA-in-Microsoft-Foundry-and-Grok-4-1-Fast-Arrives/ba-p/4497964", "supports_function_calling": true, "supports_reasoning": true, @@ -10376,11 +10389,11 @@ "max_tokens": 163840, "mode": "chat", "output_cost_per_token": 2.22e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_reasoning": true, + "supports_tool_choice": true }, "bedrock/ap-northeast-1/minimax.minimax-m2.1": { "input_cost_per_token": 3.6e-07, @@ -10390,10 +10403,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/ap-northeast-1/minimax.minimax-m2.5": { "input_cost_per_token": 3.6e-07, @@ -10402,12 +10415,12 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.44e-06, "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "output_cost_per_token": 1.44e-06 + "supports_tool_choice": true }, "bedrock/ap-northeast-1/moonshotai.kimi-k2-thinking": { "input_cost_per_token": 7.3e-07, @@ -10428,11 +10441,11 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 3.6e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_vision": true }, "bedrock/ap-northeast-1/qwen.qwen3-coder-next": { "input_cost_per_token": 6e-07, @@ -10442,10 +10455,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/moonshotai.kimi-k2-thinking": { "input_cost_per_token": 7.3e-07, @@ -10500,11 +10513,11 @@ "max_tokens": 163840, "mode": "chat", "output_cost_per_token": 2.22e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_reasoning": true, + "supports_tool_choice": true }, "bedrock/ap-south-1/minimax.minimax-m2.1": { "input_cost_per_token": 3.6e-07, @@ -10514,10 +10527,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/ap-south-1/minimax.minimax-m2.5": { "input_cost_per_token": 3.6e-07, @@ -10526,12 +10539,12 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.44e-06, "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "output_cost_per_token": 1.44e-06 + "supports_tool_choice": true }, "bedrock/ap-south-1/moonshotai.kimi-k2-thinking": { "input_cost_per_token": 7.1e-07, @@ -10552,11 +10565,11 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 3.6e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_vision": true }, "bedrock/ap-south-1/qwen.qwen3-coder-next": { "input_cost_per_token": 6e-07, @@ -10566,10 +10579,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/ap-southeast-2/minimax.minimax-m2.5": { "input_cost_per_token": 3.09e-07, @@ -10578,12 +10591,12 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.236e-06, "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "output_cost_per_token": 1.236e-06 + "supports_tool_choice": true }, "bedrock/ap-southeast-3/deepseek.v3.2": { "input_cost_per_token": 7.4e-07, @@ -10593,11 +10606,11 @@ "max_tokens": 163840, "mode": "chat", "output_cost_per_token": 2.22e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_reasoning": true, + "supports_tool_choice": true }, "bedrock/ap-southeast-3/minimax.minimax-m2.1": { "input_cost_per_token": 3.6e-07, @@ -10607,10 +10620,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/ap-southeast-3/minimax.minimax-m2.5": { "input_cost_per_token": 3.6e-07, @@ -10619,12 +10632,12 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.44e-06, "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "output_cost_per_token": 1.44e-06 + "supports_tool_choice": true }, "bedrock/ap-southeast-3/moonshotai.kimi-k2.5": { "input_cost_per_token": 7.2e-07, @@ -10634,11 +10647,11 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 3.6e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_vision": true }, "bedrock/ap-southeast-3/qwen.qwen3-coder-next": { "input_cost_per_token": 6e-07, @@ -10648,10 +10661,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/ca-central-1/meta.llama3-70b-instruct-v1:0": { "input_cost_per_token": 3.05e-06, @@ -10679,11 +10692,11 @@ "max_tokens": 163840, "mode": "chat", "output_cost_per_token": 2.22e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_reasoning": true, + "supports_tool_choice": true }, "bedrock/eu-north-1/minimax.minimax-m2.1": { "input_cost_per_token": 3.6e-07, @@ -10693,10 +10706,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/eu-north-1/minimax.minimax-m2.5": { "input_cost_per_token": 3.6e-07, @@ -10705,12 +10718,12 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.44e-06, "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "output_cost_per_token": 1.44e-06 + "supports_tool_choice": true }, "bedrock/eu-north-1/moonshotai.kimi-k2.5": { "input_cost_per_token": 7.2e-07, @@ -10720,11 +10733,11 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 3.6e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_vision": true }, "bedrock/eu-central-1/1-month-commitment/anthropic.claude-instant-v1": { "input_cost_per_second": 0.01635, @@ -10821,10 +10834,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/eu-central-1/minimax.minimax-m2.5": { "input_cost_per_token": 3.6e-07, @@ -10833,12 +10846,12 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.44e-06, "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "output_cost_per_token": 1.44e-06 + "supports_tool_choice": true }, "bedrock/eu-central-1/qwen.qwen3-coder-next": { "input_cost_per_token": 6e-07, @@ -10848,10 +10861,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/eu-west-1/meta.llama3-70b-instruct-v1:0": { "input_cost_per_token": 2.86e-06, @@ -10879,10 +10892,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/eu-west-1/minimax.minimax-m2.5": { "input_cost_per_token": 3.6e-07, @@ -10891,12 +10904,12 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.44e-06, "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "output_cost_per_token": 1.44e-06 + "supports_tool_choice": true }, "bedrock/eu-west-1/qwen.qwen3-coder-next": { "input_cost_per_token": 6e-07, @@ -10906,10 +10919,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/eu-west-2/meta.llama3-70b-instruct-v1:0": { "input_cost_per_token": 3.45e-06, @@ -10937,10 +10950,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.86e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/eu-west-2/minimax.minimax-m2.5": { "input_cost_per_token": 4.7e-07, @@ -10949,12 +10962,12 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.86e-06, "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "output_cost_per_token": 1.86e-06 + "supports_tool_choice": true }, "bedrock/eu-west-2/qwen.qwen3-coder-next": { "input_cost_per_token": 7.8e-07, @@ -10964,10 +10977,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.86e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/eu-west-3/mistral.mistral-7b-instruct-v0:2": { "input_cost_per_token": 2e-07, @@ -11007,10 +11020,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/eu-south-1/minimax.minimax-m2.5": { "input_cost_per_token": 3.6e-07, @@ -11019,12 +11032,12 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.44e-06, "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "output_cost_per_token": 1.44e-06 + "supports_tool_choice": true }, "bedrock/eu-south-1/qwen.qwen3-coder-next": { "input_cost_per_token": 6e-07, @@ -11034,12 +11047,14 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/invoke/anthropic.claude-3-5-sonnet-20240620-v1:0": { + "cache_creation_input_token_cost": 3.75e-06, + "cache_read_input_token_cost": 3e-07, "input_cost_per_token": 3e-06, "litellm_provider": "bedrock", "max_input_tokens": 200000, @@ -11053,9 +11068,7 @@ "supports_function_calling": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-07, - "cache_creation_input_token_cost": 3.75e-06 + "supports_vision": true }, "bedrock/sa-east-1/meta.llama3-70b-instruct-v1:0": { "input_cost_per_token": 4.45e-06, @@ -11083,11 +11096,11 @@ "max_tokens": 163840, "mode": "chat", "output_cost_per_token": 2.22e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_reasoning": true, + "supports_tool_choice": true }, "bedrock/sa-east-1/minimax.minimax-m2.1": { "input_cost_per_token": 3.6e-07, @@ -11097,10 +11110,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/sa-east-1/minimax.minimax-m2.5": { "input_cost_per_token": 3.6e-07, @@ -11109,12 +11122,12 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.44e-06, "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "output_cost_per_token": 1.44e-06 + "supports_tool_choice": true }, "bedrock/sa-east-1/moonshotai.kimi-k2-thinking": { "input_cost_per_token": 7.3e-07, @@ -11135,11 +11148,11 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 3.6e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_vision": true }, "bedrock/sa-east-1/qwen.qwen3-coder-next": { "input_cost_per_token": 6e-07, @@ -11149,10 +11162,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.44e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/us-east-1/1-month-commitment/anthropic.claude-instant-v1": { "input_cost_per_second": 0.011, @@ -11298,11 +11311,11 @@ "max_tokens": 163840, "mode": "chat", "output_cost_per_token": 1.85e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_reasoning": true, + "supports_tool_choice": true }, "bedrock/us-east-1/minimax.minimax-m2.1": { "input_cost_per_token": 3e-07, @@ -11312,23 +11325,23 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/us-east-1/minimax.minimax-m2.5": { "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, "litellm_provider": "bedrock", "max_input_tokens": 1000000, "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/us-east-1/moonshotai.kimi-k2-thinking": { "input_cost_per_token": 6e-07, @@ -11349,11 +11362,11 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 3e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_vision": true }, "bedrock/us-east-1/qwen.qwen3-coder-next": { "input_cost_per_token": 5e-07, @@ -11363,10 +11376,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/us-east-2/deepseek.v3.2": { "input_cost_per_token": 6.2e-07, @@ -11376,11 +11389,11 @@ "max_tokens": 163840, "mode": "chat", "output_cost_per_token": 1.85e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_reasoning": true, + "supports_tool_choice": true }, "bedrock/us-east-2/minimax.minimax-m2.1": { "input_cost_per_token": 3e-07, @@ -11390,10 +11403,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/us-east-2/minimax.minimax-m2.5": { "input_cost_per_token": 3e-07, @@ -11402,12 +11415,12 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.2e-06, "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "output_cost_per_token": 1.2e-06 + "supports_tool_choice": true }, "bedrock/us-east-2/moonshotai.kimi-k2-thinking": { "input_cost_per_token": 6e-07, @@ -11428,11 +11441,11 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 3e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_vision": true }, "bedrock/us-east-2/qwen.qwen3-coder-next": { "input_cost_per_token": 5e-07, @@ -11442,10 +11455,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/us-gov-east-1/amazon.nova-pro-v1:0": { "input_cost_per_token": 9.6e-07, @@ -11507,6 +11520,8 @@ "output_cost_per_token": 1.5e-06 }, "bedrock/us-gov-east-1/anthropic.claude-3-5-sonnet-20240620-v1:0": { + "cache_creation_input_token_cost": 4.5e-06, + "cache_read_input_token_cost": 3.6e-07, "deprecation_date": "2026-07-30", "input_cost_per_token": 3.6e-06, "litellm_provider": "bedrock", @@ -11519,11 +11534,11 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3.6e-07, - "cache_creation_input_token_cost": 4.5e-06 + "supports_vision": true }, "bedrock/us-gov-east-1/anthropic.claude-3-haiku-20240307-v1:0": { + "cache_creation_input_token_cost": 3.75e-07, + "cache_read_input_token_cost": 3e-08, "deprecation_date": "2026-09-10", "input_cost_per_token": 3e-07, "litellm_provider": "bedrock", @@ -11536,9 +11551,7 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-08, - "cache_creation_input_token_cost": 3.75e-07 + "supports_vision": true }, "bedrock/us-gov-east-1/anthropic.claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 4.5e-06, @@ -11546,24 +11559,24 @@ "cache_read_input_token_cost": 3.6e-07, "input_cost_per_token": 3.6e-06, "litellm_provider": "bedrock", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.8e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_tool_search": true, + "supports_vision": true }, "bedrock/us-gov-east-1/claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 4.5e-06, @@ -11571,24 +11584,24 @@ "cache_read_input_token_cost": 3.6e-07, "input_cost_per_token": 3.6e-06, "litellm_provider": "bedrock", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.8e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_tool_search": true, + "supports_vision": true }, "bedrock/us-gov-east-1/meta.llama3-70b-instruct-v1:0": { "input_cost_per_token": 2.65e-06, @@ -11691,6 +11704,8 @@ "supports_vision": true }, "bedrock/us-gov-west-1/anthropic.claude-3-5-sonnet-20240620-v1:0": { + "cache_creation_input_token_cost": 4.5e-06, + "cache_read_input_token_cost": 3.6e-07, "deprecation_date": "2026-07-30", "input_cost_per_token": 3.6e-06, "litellm_provider": "bedrock", @@ -11703,11 +11718,11 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3.6e-07, - "cache_creation_input_token_cost": 4.5e-06 + "supports_vision": true }, "bedrock/us-gov-west-1/anthropic.claude-3-haiku-20240307-v1:0": { + "cache_creation_input_token_cost": 3.75e-07, + "cache_read_input_token_cost": 3e-08, "deprecation_date": "2026-09-10", "input_cost_per_token": 3e-07, "litellm_provider": "bedrock", @@ -11720,9 +11735,7 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-08, - "cache_creation_input_token_cost": 3.75e-07 + "supports_vision": true }, "bedrock/us-gov-west-1/anthropic.claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 4.5e-06, @@ -11730,24 +11743,24 @@ "cache_read_input_token_cost": 3.6e-07, "input_cost_per_token": 3.6e-06, "litellm_provider": "bedrock", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.8e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_tool_search": true, + "supports_vision": true }, "bedrock/us-gov-west-1/claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 4.5e-06, @@ -11755,24 +11768,24 @@ "cache_read_input_token_cost": 3.6e-07, "input_cost_per_token": 3.6e-06, "litellm_provider": "bedrock", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.8e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_tool_search": true, + "supports_vision": true }, "bedrock/us-gov-west-1/meta.llama3-70b-instruct-v1:0": { "input_cost_per_token": 2.65e-06, @@ -11938,11 +11951,11 @@ "max_tokens": 163840, "mode": "chat", "output_cost_per_token": 1.85e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_reasoning": true, + "supports_tool_choice": true }, "bedrock/us-west-2/minimax.minimax-m2.1": { "input_cost_per_token": 3e-07, @@ -11952,23 +11965,23 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/us-west-2/minimax.minimax-m2.5": { "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, "litellm_provider": "bedrock", "max_input_tokens": 1000000, "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/us-west-2/moonshotai.kimi-k2-thinking": { "input_cost_per_token": 6e-07, @@ -11989,11 +12002,11 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 3e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_vision": true }, "bedrock/us-west-2/qwen.qwen3-coder-next": { "input_cost_per_token": 5e-07, @@ -12003,10 +12016,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/us.anthropic.claude-3-5-haiku-20241022-v1:0": { "cache_creation_input_token_cost": 1e-06, @@ -12018,13 +12031,13 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 4e-06, + "prompt_cache_min_tokens": 2048, "supports_assistant_prefill": true, "supports_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_response_schema": true, - "supports_tool_choice": true, - "prompt_cache_min_tokens": 2048 + "supports_tool_choice": true }, "black_forest_labs/flux-kontext-pro": { "litellm_provider": "black_forest_labs", @@ -12176,6 +12189,7 @@ "supports_tool_choice": true }, "cerebras/zai-glm-4.7": { + "deprecation_date": "2026-08-17", "input_cost_per_token": 2.25e-06, "litellm_provider": "cerebras", "max_input_tokens": 128000, @@ -12185,10 +12199,9 @@ "output_cost_per_token": 2.75e-06, "source": "https://www.cerebras.ai/pricing", "supports_function_calling": true, + "supports_none_reasoning_effort": true, "supports_reasoning": true, - "supports_tool_choice": true, - "deprecation_date": "2026-08-17", - "supports_none_reasoning_effort": true + "supports_tool_choice": true }, "chatdolphin": { "input_cost_per_token": 5e-07, @@ -12229,10 +12242,10 @@ ] }, "claude-haiku-4-5-20251001": { - "deprecation_date": "2026-10-15", "cache_creation_input_token_cost": 1.25e-06, "cache_creation_input_token_cost_above_1hr": 2e-06, "cache_read_input_token_cost": 1e-07, + "deprecation_date": "2026-10-15", "input_cost_per_token": 1e-06, "litellm_provider": "anthropic", "max_input_tokens": 200000, @@ -12240,23 +12253,23 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, + "prompt_cache_min_tokens": 4096, "supports_assistant_prefill": true, + "supports_computer_use": true, "supports_function_calling": true, "supports_native_structured_output": true, - "supports_computer_use": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "claude-haiku-4-5": { - "deprecation_date": "2026-10-15", "cache_creation_input_token_cost": 1.25e-06, "cache_creation_input_token_cost_above_1hr": 2e-06, "cache_read_input_token_cost": 1e-07, + "deprecation_date": "2026-10-15", "input_cost_per_token": 1e-06, "litellm_provider": "anthropic", "max_input_tokens": 200000, @@ -12264,17 +12277,17 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, + "prompt_cache_min_tokens": 4096, "supports_assistant_prefill": true, + "supports_computer_use": true, "supports_function_calling": true, "supports_native_structured_output": true, - "supports_computer_use": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "claude-3-7-sonnet-20250219": { "cache_creation_input_token_cost": 3.75e-06, @@ -12353,6 +12366,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12366,8 +12380,7 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "claude-4-sonnet-20250514": { "cache_creation_input_token_cost": 3.75e-06, @@ -12384,6 +12397,7 @@ "mode": "chat", "output_cost_per_token": 1.5e-05, "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12398,26 +12412,26 @@ "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "prompt_cache_min_tokens": 1024 + "supports_web_search": true }, "claude-sonnet-4-5": { - "deprecation_date": "2026-09-29", "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.2e-05, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "deprecation_date": "2026-09-29", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "litellm_provider": "anthropic", "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12426,32 +12440,32 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "claude-sonnet-4-5-20250929": { - "deprecation_date": "2026-09-29", "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.2e-05, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "deprecation_date": "2026-09-29", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "litellm_provider": "anthropic", "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12460,21 +12474,20 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "prompt_cache_min_tokens": 1024 + "supports_web_search": true }, "claude-sonnet-5": { - "deprecation_date": "2027-06-30", "cache_creation_input_token_cost": 2.5e-06, "cache_creation_input_token_cost_above_1hr": 4e-06, "cache_read_input_token_cost": 2e-07, + "deprecation_date": "2027-06-30", "input_cost_per_token": 2e-06, "litellm_provider": "anthropic", "max_input_tokens": 1000000, @@ -12482,17 +12495,23 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1e-05, + "prompt_cache_min_tokens": 1024, + "provider_specific_entry": { + "us": 1.1 + }, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_native_structured_output": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -12500,19 +12519,13 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "provider_specific_entry": { - "us": 1.1 - }, - "supports_output_config": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "claude-sonnet-4-6": { - "deprecation_date": "2027-02-17", "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, "cache_read_input_token_cost": 3e-07, + "deprecation_date": "2027-02-17", "input_cost_per_token": 3e-06, "litellm_provider": "anthropic", "max_input_tokens": 1000000, @@ -12520,6 +12533,7 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12529,52 +12543,53 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.2e-05, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "input_cost_per_token_batches": 1.5e-06, "litellm_provider": "bedrock", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "output_cost_per_token_batches": 7.5e-06, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024, - "input_cost_per_token_batches": 1.5e-06, - "output_cost_per_token_batches": 7.5e-06 + "supports_tool_search": true, + "supports_vision": true }, "claude-opus-4-1": { "cache_creation_input_token_cost": 1.875e-05, "cache_creation_input_token_cost_above_1hr": 3e-05, "cache_read_input_token_cost": 1.5e-06, + "deprecation_date": "2026-08-05", "input_cost_per_token": 1.5e-05, "litellm_provider": "anthropic", "max_input_tokens": 200000, @@ -12582,6 +12597,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12590,28 +12606,27 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024, - "deprecation_date": "2026-08-05" + "supports_vision": true }, "claude-opus-4-1-20250805": { "cache_creation_input_token_cost": 1.875e-05, "cache_creation_input_token_cost_above_1hr": 3e-05, "cache_read_input_token_cost": 1.5e-06, - "input_cost_per_token": 1.5e-05, "deprecation_date": "2026-08-05", + "input_cost_per_token": 1.5e-05, "litellm_provider": "anthropic", "max_input_tokens": 200000, "max_output_tokens": 32000, "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12620,27 +12635,27 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "claude-opus-4-20250514": { "cache_creation_input_token_cost": 1.875e-05, "cache_creation_input_token_cost_above_1hr": 3e-05, "cache_read_input_token_cost": 1.5e-06, - "input_cost_per_token": 1.5e-05, "deprecation_date": "2026-06-15", + "input_cost_per_token": 1.5e-05, "litellm_provider": "anthropic", "max_input_tokens": 200000, "max_output_tokens": 32000, "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12654,14 +12669,13 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "claude-opus-4-5-20251101": { - "deprecation_date": "2026-11-24", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2026-11-24", "input_cost_per_token": 5e-06, "litellm_provider": "anthropic", "max_input_tokens": 200000, @@ -12669,6 +12683,7 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12677,21 +12692,20 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "claude-opus-4-5": { - "deprecation_date": "2026-11-24", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2026-11-24", "input_cost_per_token": 5e-06, "litellm_provider": "anthropic", "max_input_tokens": 200000, @@ -12699,6 +12713,7 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12707,21 +12722,20 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "claude-opus-4-6": { - "deprecation_date": "2027-02-05", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-02-05", "input_cost_per_token": 5e-06, "litellm_provider": "anthropic", "max_input_tokens": 1000000, @@ -12729,6 +12743,11 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, + "provider_specific_entry": { + "us": 1.1, + "fast": 6.0 + }, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12738,27 +12757,23 @@ "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, - "supports_tool_choice": true, - "supports_vision": true, - "provider_specific_entry": { - "us": 1.1, - "fast": 6.0 - }, - "supports_output_config": true, - "supports_max_reasoning_effort": true, "supports_speed": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_choice": true, + "supports_vision": true }, "claude-opus-4-6-20260205": { - "deprecation_date": "2027-02-05", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-02-05", "input_cost_per_token": 5e-06, "litellm_provider": "anthropic", "max_input_tokens": 1000000, @@ -12766,6 +12781,11 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, + "provider_specific_entry": { + "us": 1.1, + "fast": 6.0 + }, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12775,27 +12795,23 @@ "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, - "supports_tool_choice": true, - "supports_vision": true, - "provider_specific_entry": { - "us": 1.1, - "fast": 6.0 - }, - "supports_max_reasoning_effort": true, - "supports_output_config": true, "supports_speed": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_choice": true, + "supports_vision": true }, "claude-opus-4-7": { - "deprecation_date": "2027-04-16", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-04-16", "input_cost_per_token": 5e-06, "litellm_provider": "anthropic", "max_input_tokens": 1000000, @@ -12803,6 +12819,11 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 2048, + "provider_specific_entry": { + "us": 1.1, + "fast": 6.0 + }, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12812,29 +12833,24 @@ "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, "supports_sampling_params": false, + "supports_speed": true, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "provider_specific_entry": { - "us": 1.1, - "fast": 6.0 - }, - "supports_output_config": true, - "supports_speed": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "claude-opus-4-7-20260416": { - "deprecation_date": "2027-04-16", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-04-16", "input_cost_per_token": 5e-06, "litellm_provider": "anthropic", "max_input_tokens": 1000000, @@ -12842,6 +12858,11 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 2048, + "provider_specific_entry": { + "us": 1.1, + "fast": 6.0 + }, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -12851,29 +12872,24 @@ "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, "supports_sampling_params": false, + "supports_speed": true, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "provider_specific_entry": { - "us": 1.1, - "fast": 6.0 - }, - "supports_output_config": true, - "supports_speed": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "claude-fable-5": { - "deprecation_date": "2027-06-09", "cache_creation_input_token_cost": 1.25e-05, "cache_creation_input_token_cost_above_1hr": 2e-05, "cache_read_input_token_cost": 1e-06, + "deprecation_date": "2027-06-09", "input_cost_per_token": 1e-05, "litellm_provider": "anthropic", "max_input_tokens": 1000000, @@ -12881,38 +12897,38 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 5e-05, - "search_context_cost_per_query": { - "search_context_size_high": 0.01, - "search_context_size_low": 0.01, - "search_context_size_medium": 0.01 - }, - "supports_adaptive_thinking": true, - "thinking_always_on": true, - "supports_mid_conversation_system": true, - "supports_assistant_prefill": false, - "supports_computer_use": true, - "supports_function_calling": true, - "supports_pdf_input": true, - "supports_prompt_caching": true, - "supports_reasoning": true, - "supports_response_schema": true, - "supports_sampling_params": false, - "supports_tool_choice": true, - "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, + "prompt_cache_min_tokens": 512, "provider_specific_entry": { "us": 1.1 }, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, + "supports_adaptive_thinking": true, + "supports_assistant_prefill": false, + "supports_computer_use": true, + "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, "supports_output_config": true, - "prompt_cache_min_tokens": 512, - "supports_native_structured_output": true + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_sampling_params": false, + "supports_tool_choice": true, + "supports_vision": true, + "supports_xhigh_reasoning_effort": true, + "thinking_always_on": true }, "claude-opus-5": { - "deprecation_date": "2027-07-24", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-07-24", "input_cost_per_token": 5e-06, "litellm_provider": "anthropic", "max_input_tokens": 1000000, @@ -12920,39 +12936,39 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 512, + "provider_specific_entry": { + "us": 1.1, + "fast": 2.0 + }, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, "supports_sampling_params": false, + "supports_speed": true, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "provider_specific_entry": { - "us": 1.1, - "fast": 2.0 - }, - "supports_output_config": true, - "supports_speed": true, - "prompt_cache_min_tokens": 512 + "supports_xhigh_reasoning_effort": true }, "claude-opus-4-8": { - "deprecation_date": "2027-05-28", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-05-28", "input_cost_per_token": 5e-06, "litellm_provider": "anthropic", "max_input_tokens": 1000000, @@ -12960,50 +12976,51 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 1024, + "provider_specific_entry": { + "us": 1.1, + "fast": 2.0 + }, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_native_structured_output": true, "supports_sampling_params": false, + "supports_speed": true, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "provider_specific_entry": { - "us": 1.1, - "fast": 2.0 - }, - "supports_output_config": true, - "supports_speed": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "claude-sonnet-4-20250514": { - "deprecation_date": "2026-06-15", "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "deprecation_date": "2026-06-15", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "litellm_provider": "anthropic", "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -13017,8 +13034,7 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "cloudflare/@cf/meta/llama-2-7b-chat-fp16": { "input_cost_per_token": 1.923e-06, @@ -13463,14 +13479,14 @@ "output_cost_per_token": 0.0 }, "command": { + "deprecation_date": "2025-09-15", "input_cost_per_token": 1e-06, "litellm_provider": "cohere", "max_input_tokens": 4096, "max_output_tokens": 4096, "max_tokens": 4096, "mode": "completion", - "output_cost_per_token": 2e-06, - "deprecation_date": "2025-09-15" + "output_cost_per_token": 2e-06 }, "command-a-03-2025": { "input_cost_per_token": 2.5e-06, @@ -13484,6 +13500,7 @@ "supports_tool_choice": true }, "command-light": { + "deprecation_date": "2025-09-15", "input_cost_per_token": 3e-07, "litellm_provider": "cohere_chat", "max_input_tokens": 4096, @@ -13491,8 +13508,7 @@ "max_tokens": 4096, "mode": "chat", "output_cost_per_token": 6e-07, - "supports_tool_choice": true, - "deprecation_date": "2025-09-15" + "supports_tool_choice": true }, "command-nightly": { "input_cost_per_token": 1e-06, @@ -13504,6 +13520,7 @@ "output_cost_per_token": 2e-06 }, "command-r": { + "deprecation_date": "2025-09-15", "input_cost_per_token": 1.5e-07, "litellm_provider": "cohere_chat", "max_input_tokens": 128000, @@ -13512,8 +13529,7 @@ "mode": "chat", "output_cost_per_token": 6e-07, "supports_function_calling": true, - "supports_tool_choice": true, - "deprecation_date": "2025-09-15" + "supports_tool_choice": true }, "command-r-08-2024": { "input_cost_per_token": 1.5e-07, @@ -13527,6 +13543,7 @@ "supports_tool_choice": true }, "command-r-plus": { + "deprecation_date": "2025-09-15", "input_cost_per_token": 2.5e-06, "litellm_provider": "cohere_chat", "max_input_tokens": 128000, @@ -13535,8 +13552,7 @@ "mode": "chat", "output_cost_per_token": 1e-05, "supports_function_calling": true, - "supports_tool_choice": true, - "deprecation_date": "2025-09-15" + "supports_tool_choice": true }, "command-r-plus-08-2024": { "input_cost_per_token": 2.5e-06, @@ -14674,8 +14690,8 @@ "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "databricks/databricks-claude-opus-4": { "cache_creation_input_token_cost": 1.874999e-05, @@ -14719,8 +14735,8 @@ "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "databricks/databricks-claude-opus-4-5": { "cache_creation_input_token_cost": 6.25002e-06, @@ -14740,11 +14756,11 @@ "source": "https://www.databricks.com/product/pricing/proprietary-foundation-model-serving", "supports_assistant_prefill": true, "supports_function_calling": true, + "supports_output_config": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "supports_output_config": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "databricks/databricks-claude-opus-4-6": { "cache_creation_input_token_cost": 6.25002e-06, @@ -14764,6 +14780,7 @@ "source": "https://www.databricks.com/product/pricing/proprietary-foundation-model-serving", "supports_assistant_prefill": true, "supports_function_calling": true, + "supports_legacy_thinking": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_tool_choice": true @@ -14868,8 +14885,8 @@ "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "databricks/databricks-claude-sonnet-4-1": { "cache_creation_input_token_cost": 3.74997e-06, @@ -14913,8 +14930,8 @@ "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "databricks/databricks-claude-sonnet-4-6": { "cache_creation_input_token_cost": 3.74997e-06, @@ -14934,6 +14951,7 @@ "source": "https://www.databricks.com/product/pricing/proprietary-foundation-model-serving", "supports_assistant_prefill": true, "supports_function_calling": true, + "supports_legacy_thinking": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_tool_choice": true @@ -14983,8 +15001,8 @@ "source": "https://www.databricks.com/product/pricing/proprietary-foundation-model-serving", "supports_function_calling": true, "supports_prompt_caching": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "databricks/databricks-gemini-2-5-pro": { "cache_creation_input_token_cost": 1.24999e-06, @@ -15004,8 +15022,8 @@ "source": "https://www.databricks.com/product/pricing/proprietary-foundation-model-serving", "supports_function_calling": true, "supports_prompt_caching": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "databricks/databricks-gemini-3-1-flash-lite": { "cache_creation_input_token_cost": 3.1248e-07, @@ -15412,8 +15430,8 @@ "output_cost_per_token": 1.5000300000000002e-06, "output_dbu_cost_per_token": 2.1429e-05, "source": "https://www.databricks.com/product/pricing/foundation-model-serving", - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "databricks/databricks-meta-llama-3-1-405b-instruct": { "cache_creation_input_token_cost": 5.00003e-06, @@ -15467,8 +15485,8 @@ "output_cost_per_token": 1.5000300000000002e-06, "output_dbu_cost_per_token": 2.1429e-05, "source": "https://www.databricks.com/product/pricing/foundation-model-serving", - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "databricks/databricks-meta-llama-3-70b-instruct": { "cache_creation_input_token_cost": 1.00002e-06, @@ -16056,750 +16074,750 @@ ] }, "deepinfra/Gryphe/MythoMax-L2-13b": { - "max_tokens": 4096, + "input_cost_per_token": 8e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 8e-08, - "output_cost_per_token": 9e-08, - "litellm_provider": "deepinfra", + "max_tokens": 4096, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 9e-08, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/NousResearch/Hermes-3-Llama-3.1-405B": { - "max_tokens": 131072, + "input_cost_per_token": 1e-06, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-06, - "output_cost_per_token": 1e-06, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/NousResearch/Hermes-3-Llama-3.1-70B": { - "max_tokens": 131072, + "input_cost_per_token": 3e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 3e-07, - "output_cost_per_token": 3e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 3e-07, "supports_tool_choice": false }, "deepinfra/Qwen/QwQ-32B": { - "max_tokens": 131072, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 4e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Qwen/Qwen2.5-72B-Instruct": { - "max_tokens": 32768, + "input_cost_per_token": 1.2e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1.2e-07, - "output_cost_per_token": 3.9e-07, - "litellm_provider": "deepinfra", + "max_tokens": 32768, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 3.9e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Qwen/Qwen2.5-7B-Instruct": { - "max_tokens": 32768, + "input_cost_per_token": 4e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 4e-08, - "output_cost_per_token": 1e-07, - "litellm_provider": "deepinfra", + "max_tokens": 32768, "mode": "chat", + "output_cost_per_token": 1e-07, "supports_tool_choice": false }, "deepinfra/Qwen/Qwen2.5-VL-32B-Instruct": { - "max_tokens": 128000, + "input_cost_per_token": 2e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "deepinfra", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 6e-07, + "supports_function_calling": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_function_calling": true + "supports_vision": true }, "deepinfra/Qwen/Qwen3-14B": { - "max_tokens": 40960, + "input_cost_per_token": 6e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 6e-08, - "output_cost_per_token": 2.4e-07, - "litellm_provider": "deepinfra", + "max_tokens": 40960, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2.4e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Qwen/Qwen3-235B-A22B": { - "max_tokens": 40960, + "input_cost_per_token": 1.8e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 1.8e-07, - "output_cost_per_token": 5.4e-07, - "litellm_provider": "deepinfra", + "max_tokens": 40960, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 5.4e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Qwen/Qwen3-235B-A22B-Instruct-2507": { - "max_tokens": 262144, + "input_cost_per_token": 9e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 9e-08, - "output_cost_per_token": 6e-07, - "litellm_provider": "deepinfra", + "max_tokens": 262144, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 6e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Qwen/Qwen3-235B-A22B-Thinking-2507": { - "max_tokens": 262144, + "input_cost_per_token": 3e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 3e-07, - "output_cost_per_token": 2.9e-06, - "litellm_provider": "deepinfra", + "max_tokens": 262144, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2.9e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Qwen/Qwen3-30B-A3B": { - "max_tokens": 40960, + "input_cost_per_token": 8e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 8e-08, - "output_cost_per_token": 2.9e-07, - "litellm_provider": "deepinfra", + "max_tokens": 40960, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2.9e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Qwen/Qwen3-32B": { - "max_tokens": 40960, + "input_cost_per_token": 1e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 2.8e-07, - "litellm_provider": "deepinfra", + "max_tokens": 40960, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2.8e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Qwen/Qwen3-Coder-480B-A35B-Instruct": { - "max_tokens": 262144, + "input_cost_per_token": 4e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 4e-07, - "output_cost_per_token": 1.6e-06, - "litellm_provider": "deepinfra", + "max_tokens": 262144, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1.6e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Qwen/Qwen3-Coder-480B-A35B-Instruct-Turbo": { - "max_tokens": 262144, + "input_cost_per_token": 2.9e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 2.9e-07, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "deepinfra", + "max_tokens": 262144, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1.2e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Qwen/Qwen3-Next-80B-A3B-Instruct": { - "max_tokens": 262144, + "input_cost_per_token": 1.4e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1.4e-07, - "output_cost_per_token": 1.4e-06, - "litellm_provider": "deepinfra", + "max_tokens": 262144, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1.4e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Qwen/Qwen3-Next-80B-A3B-Thinking": { - "max_tokens": 262144, + "input_cost_per_token": 1.4e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1.4e-07, - "output_cost_per_token": 1.4e-06, - "litellm_provider": "deepinfra", + "max_tokens": 262144, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1.4e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/Sao10K/L3-8B-Lunaris-v1-Turbo": { - "max_tokens": 8192, + "input_cost_per_token": 4e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 4e-08, - "output_cost_per_token": 5e-08, - "litellm_provider": "deepinfra", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 5e-08, "supports_tool_choice": false }, "deepinfra/Sao10K/L3.1-70B-Euryale-v2.2": { - "max_tokens": 131072, + "input_cost_per_token": 6.5e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 6.5e-07, - "output_cost_per_token": 7.5e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 7.5e-07, "supports_tool_choice": false }, "deepinfra/Sao10K/L3.3-70B-Euryale-v2.3": { - "max_tokens": 131072, + "input_cost_per_token": 6.5e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 6.5e-07, - "output_cost_per_token": 7.5e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 7.5e-07, "supports_tool_choice": false }, "deepinfra/allenai/olmOCR-7B-0725-FP8": { - "max_tokens": 16384, + "input_cost_per_token": 2.7e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 2.7e-07, - "output_cost_per_token": 1.5e-06, - "litellm_provider": "deepinfra", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 1.5e-06, "supports_tool_choice": false }, "deepinfra/anthropic/claude-3-7-sonnet-latest": { - "max_tokens": 200000, + "cache_read_input_token_cost": 3.3e-07, + "input_cost_per_token": 3.3e-06, + "litellm_provider": "deepinfra", "max_input_tokens": 200000, "max_output_tokens": 200000, - "input_cost_per_token": 3.3e-06, - "output_cost_per_token": 1.65e-05, - "cache_read_input_token_cost": 3.3e-07, - "litellm_provider": "deepinfra", + "max_tokens": 200000, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1.65e-05, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/anthropic/claude-4-opus": { - "max_tokens": 200000, + "input_cost_per_token": 1.65e-05, + "litellm_provider": "deepinfra", "max_input_tokens": 200000, "max_output_tokens": 200000, - "input_cost_per_token": 1.65e-05, - "output_cost_per_token": 8.25e-05, - "litellm_provider": "deepinfra", + "max_tokens": 200000, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 8.25e-05, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/anthropic/claude-4-sonnet": { - "max_tokens": 200000, + "input_cost_per_token": 3.3e-06, + "litellm_provider": "deepinfra", "max_input_tokens": 200000, "max_output_tokens": 200000, - "input_cost_per_token": 3.3e-06, - "output_cost_per_token": 1.65e-05, - "litellm_provider": "deepinfra", + "max_tokens": 200000, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1.65e-05, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/deepseek-ai/DeepSeek-R1": { - "max_tokens": 163840, + "input_cost_per_token": 7e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 7e-07, - "output_cost_per_token": 2.4e-06, - "litellm_provider": "deepinfra", + "max_tokens": 163840, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2.4e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/deepseek-ai/DeepSeek-R1-0528": { - "max_tokens": 163840, + "cache_read_input_token_cost": 4e-07, + "input_cost_per_token": 5e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 2.15e-06, - "cache_read_input_token_cost": 4e-07, - "litellm_provider": "deepinfra", + "max_tokens": 163840, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2.15e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/deepseek-ai/DeepSeek-R1-0528-Turbo": { - "max_tokens": 32768, + "input_cost_per_token": 1e-06, + "litellm_provider": "deepinfra", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-06, - "output_cost_per_token": 3e-06, - "litellm_provider": "deepinfra", + "max_tokens": 32768, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 3e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/deepseek-ai/DeepSeek-R1-Distill-Llama-70B": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 6e-07, "supports_tool_choice": false }, "deepinfra/deepseek-ai/DeepSeek-R1-Distill-Qwen-32B": { - "max_tokens": 131072, + "input_cost_per_token": 2.7e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2.7e-07, - "output_cost_per_token": 2.7e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2.7e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/deepseek-ai/DeepSeek-R1-Turbo": { - "max_tokens": 40960, + "input_cost_per_token": 1e-06, + "litellm_provider": "deepinfra", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 1e-06, - "output_cost_per_token": 3e-06, - "litellm_provider": "deepinfra", + "max_tokens": 40960, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 3e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/deepseek-ai/DeepSeek-V3": { - "max_tokens": 163840, + "input_cost_per_token": 3.8e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 3.8e-07, - "output_cost_per_token": 8.9e-07, - "litellm_provider": "deepinfra", + "max_tokens": 163840, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 8.9e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/deepseek-ai/DeepSeek-V3-0324": { - "max_tokens": 163840, + "input_cost_per_token": 2.5e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 2.5e-07, - "output_cost_per_token": 8.8e-07, - "litellm_provider": "deepinfra", + "max_tokens": 163840, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 8.8e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/deepseek-ai/DeepSeek-V3.1": { - "max_tokens": 163840, + "cache_read_input_token_cost": 2.16e-07, + "input_cost_per_token": 2.7e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 2.7e-07, - "output_cost_per_token": 1e-06, - "cache_read_input_token_cost": 2.16e-07, - "litellm_provider": "deepinfra", + "max_tokens": 163840, "mode": "chat", - "supports_tool_choice": true, + "output_cost_per_token": 1e-06, + "supports_function_calling": true, "supports_reasoning": true, - "supports_function_calling": true + "supports_tool_choice": true }, "deepinfra/deepseek-ai/DeepSeek-V3.1-Terminus": { - "max_tokens": 163840, + "cache_read_input_token_cost": 2.16e-07, + "input_cost_per_token": 2.7e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 2.7e-07, - "output_cost_per_token": 1e-06, - "cache_read_input_token_cost": 2.16e-07, - "litellm_provider": "deepinfra", + "max_tokens": 163840, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/google/gemini-2.0-flash-001": { "deprecation_date": "2026-06-01", - "max_tokens": 1000000, + "input_cost_per_token": 1e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 1000000, "max_output_tokens": 1000000, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "deepinfra", + "max_tokens": 1000000, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 4e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/google/gemini-2.5-flash": { - "max_tokens": 1000000, + "input_cost_per_token": 3e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 1000000, "max_output_tokens": 1000000, - "input_cost_per_token": 3e-07, - "output_cost_per_token": 2.5e-06, - "litellm_provider": "deepinfra", + "max_tokens": 1000000, "mode": "chat", - "supports_tool_choice": true, + "output_cost_per_token": 2.5e-06, "supports_function_calling": true, - "supports_image_size": false + "supports_image_size": false, + "supports_tool_choice": true }, "deepinfra/google/gemini-2.5-pro": { - "max_tokens": 1000000, + "input_cost_per_token": 1.25e-06, + "litellm_provider": "deepinfra", "max_input_tokens": 1000000, "max_output_tokens": 1000000, - "input_cost_per_token": 1.25e-06, - "output_cost_per_token": 1e-05, - "litellm_provider": "deepinfra", + "max_tokens": 1000000, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1e-05, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/google/gemma-3-12b-it": { - "max_tokens": 131072, + "input_cost_per_token": 5e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 5e-08, - "output_cost_per_token": 1e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/google/gemma-3-27b-it": { - "max_tokens": 131072, + "input_cost_per_token": 9e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-08, - "output_cost_per_token": 1.6e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1.6e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/google/gemma-3-4b-it": { - "max_tokens": 131072, + "input_cost_per_token": 4e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 4e-08, - "output_cost_per_token": 8e-08, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 8e-08, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/meta-llama/Llama-3.2-11B-Vision-Instruct": { - "max_tokens": 131072, + "input_cost_per_token": 4.9e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 4.9e-08, - "output_cost_per_token": 4.9e-08, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 4.9e-08, "supports_tool_choice": false }, "deepinfra/meta-llama/Llama-3.2-3B-Instruct": { - "max_tokens": 131072, + "input_cost_per_token": 2e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-08, - "output_cost_per_token": 2e-08, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2e-08, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/meta-llama/Llama-3.3-70B-Instruct": { - "max_tokens": 131072, + "input_cost_per_token": 2.3e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2.3e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 4e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/meta-llama/Llama-3.3-70B-Instruct-Turbo": { - "max_tokens": 131072, + "input_cost_per_token": 1.3e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1.3e-07, - "output_cost_per_token": 3.9e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 3.9e-07, "supports_function_calling": true, "supports_tool_choice": true }, "deepinfra/meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8": { - "max_tokens": 1048576, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 1048576, "max_output_tokens": 1048576, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "deepinfra", + "max_tokens": 1048576, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 6e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/meta-llama/Llama-4-Scout-17B-16E-Instruct": { - "max_tokens": 327680, + "input_cost_per_token": 8e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 327680, "max_output_tokens": 327680, - "input_cost_per_token": 8e-08, - "output_cost_per_token": 3e-07, - "litellm_provider": "deepinfra", + "max_tokens": 327680, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 3e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/meta-llama/Llama-Guard-3-8B": { - "max_tokens": 131072, + "input_cost_per_token": 5.5e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 5.5e-08, - "output_cost_per_token": 5.5e-08, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 5.5e-08, "supports_tool_choice": false }, "deepinfra/meta-llama/Llama-Guard-4-12B": { - "max_tokens": 163840, + "input_cost_per_token": 1.8e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 1.8e-07, - "output_cost_per_token": 1.8e-07, - "litellm_provider": "deepinfra", + "max_tokens": 163840, "mode": "chat", + "output_cost_per_token": 1.8e-07, "supports_tool_choice": false }, "deepinfra/meta-llama/Meta-Llama-3-8B-Instruct": { - "max_tokens": 8192, + "input_cost_per_token": 3e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 3e-08, - "output_cost_per_token": 6e-08, - "litellm_provider": "deepinfra", + "max_tokens": 8192, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 6e-08, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/meta-llama/Meta-Llama-3.1-70B-Instruct": { - "max_tokens": 131072, + "input_cost_per_token": 4e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 4e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 4e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo": { - "max_tokens": 131072, + "input_cost_per_token": 1e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 2.8e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2.8e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/meta-llama/Meta-Llama-3.1-8B-Instruct": { - "max_tokens": 131072, + "input_cost_per_token": 3e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 3e-08, - "output_cost_per_token": 5e-08, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 5e-08, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo": { - "max_tokens": 131072, + "input_cost_per_token": 2e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-08, - "output_cost_per_token": 3e-08, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 3e-08, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/microsoft/WizardLM-2-8x22B": { - "max_tokens": 65536, + "input_cost_per_token": 4.8e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 65536, "max_output_tokens": 65536, - "input_cost_per_token": 4.8e-07, - "output_cost_per_token": 4.8e-07, - "litellm_provider": "deepinfra", + "max_tokens": 65536, "mode": "chat", + "output_cost_per_token": 4.8e-07, "supports_tool_choice": false }, "deepinfra/microsoft/phi-4": { - "max_tokens": 16384, + "input_cost_per_token": 7e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 7e-08, - "output_cost_per_token": 1.4e-07, - "litellm_provider": "deepinfra", + "max_tokens": 16384, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1.4e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/mistralai/Mistral-Nemo-Instruct-2407": { - "max_tokens": 131072, + "input_cost_per_token": 2e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-08, - "output_cost_per_token": 4e-08, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 4e-08, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/mistralai/Mistral-Small-24B-Instruct-2501": { - "max_tokens": 32768, + "input_cost_per_token": 5e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 5e-08, - "output_cost_per_token": 8e-08, - "litellm_provider": "deepinfra", + "max_tokens": 32768, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 8e-08, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/mistralai/Mistral-Small-3.2-24B-Instruct-2506": { - "max_tokens": 128000, + "input_cost_per_token": 7.5e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 7.5e-08, - "output_cost_per_token": 2e-07, - "litellm_provider": "deepinfra", + "max_tokens": 128000, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/mistralai/Mixtral-8x7B-Instruct-v0.1": { - "max_tokens": 32768, + "input_cost_per_token": 4e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 4e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "deepinfra", + "max_tokens": 32768, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 4e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/moonshotai/Kimi-K2-Instruct": { - "max_tokens": 131072, + "input_cost_per_token": 5e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 2e-06, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/moonshotai/Kimi-K2-Instruct-0905": { - "max_tokens": 262144, + "cache_read_input_token_cost": 4e-07, + "input_cost_per_token": 5e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 2e-06, - "cache_read_input_token_cost": 4e-07, - "litellm_provider": "deepinfra", + "max_tokens": 262144, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 2e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/nvidia/Llama-3.1-Nemotron-70B-Instruct": { - "max_tokens": 131072, + "input_cost_per_token": 6e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 6e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 6e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/nvidia/Llama-3.3-Nemotron-Super-49B-v1.5": { - "max_tokens": 131072, + "input_cost_per_token": 1e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 4e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/nvidia/NVIDIA-Nemotron-3.5-Lightning": { - "max_input_tokens": 262144, "input_cost_per_token": 5e-08, - "output_cost_per_token": 2e-07, "litellm_provider": "deepinfra", + "max_input_tokens": 262144, "mode": "chat", + "output_cost_per_token": 2e-07, "source": "https://deepinfra.com/nvidia/NVIDIA-Nemotron-3.5-Lightning", - "supports_tool_choice": true, "supports_function_calling": true, - "supports_reasoning": true + "supports_reasoning": true, + "supports_tool_choice": true }, "deepinfra/nvidia/NVIDIA-Nemotron-Nano-9B-v2": { - "max_tokens": 131072, + "input_cost_per_token": 4e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 4e-08, - "output_cost_per_token": 1.6e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1.6e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/openai/gpt-oss-120b": { - "max_tokens": 131072, + "input_cost_per_token": 5e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 5e-08, - "output_cost_per_token": 4.5e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 4.5e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/openai/gpt-oss-20b": { - "max_tokens": 131072, + "input_cost_per_token": 4e-08, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 4e-08, - "output_cost_per_token": 1.5e-07, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1.5e-07, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepinfra/zai-org/GLM-4.5": { - "max_tokens": 131072, + "input_cost_per_token": 4e-07, + "litellm_provider": "deepinfra", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 4e-07, - "output_cost_per_token": 1.6e-06, - "litellm_provider": "deepinfra", + "max_tokens": 131072, "mode": "chat", - "supports_tool_choice": true, - "supports_function_calling": true + "output_cost_per_token": 1.6e-06, + "supports_function_calling": true, + "supports_tool_choice": true }, "deepseek/deepseek-chat": { "cache_creation_input_token_cost": 0.0, @@ -16918,9 +16936,9 @@ "mode": "chat", "output_cost_per_token": 1.68e-06, "supports_function_calling": true, + "supports_native_structured_output": true, "supports_reasoning": true, - "supports_tool_choice": true, - "supports_native_structured_output": true + "supports_tool_choice": true }, "deepseek.v3.2": { "input_cost_per_token": 6.2e-07, @@ -16930,11 +16948,11 @@ "max_tokens": 163840, "mode": "chat", "output_cost_per_token": 1.85e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_reasoning": true, + "supports_tool_choice": true }, "dolphin": { "input_cost_per_token": 5e-07, @@ -17069,6 +17087,9 @@ }, "firecrawl/search": { "litellm_provider": "firecrawl", + "metadata": { + "notes": "Firecrawl search pricing: $83 for 100,000 credits, 2 credits per 10 results. Cost = ceiling(limit/10) * 2 * $0.00083" + }, "mode": "search", "tiered_pricing": [ { @@ -17141,10 +17162,7 @@ 100 ] } - ], - "metadata": { - "notes": "Firecrawl search pricing: $83 for 100,000 credits, 2 credits per 10 results. Cost = ceiling(limit/10) * 2 * $0.00083" - } + ] }, "perplexity/search": { "input_cost_per_query": 0.005, @@ -17152,60 +17170,60 @@ "mode": "search" }, "searxng/search": { - "litellm_provider": "searxng", - "mode": "search", "input_cost_per_query": 0.0, + "litellm_provider": "searxng", "metadata": { "notes": "SearXNG is an open-source metasearch engine. Free to use when self-hosted or using public instances." - } + }, + "mode": "search" }, "serper/search": { "input_cost_per_query": 0.001, "litellm_provider": "serper", - "mode": "search", "metadata": { "notes": "Serper Google Search API. Pricing: $1.00/1k queries (Starter), $0.75/1k (Standard), $0.50/1k (Scale), $0.30/1k (Ultimate)." - } + }, + "mode": "search" }, "apiserpent/search": { "input_cost_per_query": 0.0006, "litellm_provider": "apiserpent", - "mode": "search", "metadata": { "notes": "APISerpent quick search (/api/search/quick), multi-engine (Google, Bing, Yahoo, DuckDuckGo). Pricing: $0.60/1k searches." - } + }, + "mode": "search" }, "apiserpent/deep_search": { "input_cost_per_query": 0.0006, "litellm_provider": "apiserpent", - "mode": "search", "metadata": { "notes": "APISerpent deep search (/api/search), multi-engine (Google, Bing, Yahoo, DuckDuckGo). Pricing: $0.60/1k searches." - } + }, + "mode": "search" }, "agentcore/search": { "input_cost_per_query": 0.0, "litellm_provider": "agentcore", - "mode": "search", "metadata": { "notes": "Web Search on Amazon Bedrock AgentCore, billed by AWS on the gateway" - } + }, + "mode": "search" }, "tinyfish/search": { "input_cost_per_query": 0.0, "litellm_provider": "tinyfish", - "mode": "search", "metadata": { "notes": "TinyFish Search API" - } + }, + "mode": "search" }, "nimble/search": { "input_cost_per_query": 0.005, "litellm_provider": "nimble", - "mode": "search", "metadata": { "notes": "Nimble Search API pay-as-you-go list price: $5 per 1,000 searches, up to 100 results per search. Volume plans price differently." - } + }, + "mode": "search" }, "elevenlabs/scribe_v1": { "input_cost_per_second": 6.11e-05, @@ -17372,6 +17390,8 @@ "supports_vision": true }, "eu.anthropic.claude-3-5-haiku-20241022-v1:0": { + "cache_creation_input_token_cost": 3.125e-07, + "cache_read_input_token_cost": 2.5e-08, "input_cost_per_token": 2.5e-07, "litellm_provider": "bedrock", "max_input_tokens": 200000, @@ -17379,46 +17399,46 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.25e-06, + "prompt_cache_min_tokens": 2048, "supports_assistant_prefill": true, "supports_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_response_schema": true, - "supports_tool_choice": true, - "cache_read_input_token_cost": 2.5e-08, - "cache_creation_input_token_cost": 3.125e-07, - "prompt_cache_min_tokens": 2048 + "supports_tool_choice": true }, "eu.anthropic.claude-haiku-4-5-20251001-v1:0": { "cache_creation_input_token_cost": 1.375e-06, "cache_creation_input_token_cost_above_1hr": 2.2e-06, "cache_read_input_token_cost": 1.1e-07, - "input_cost_per_token": 1.1e-06, "deprecation_date": "2026-10-15", + "input_cost_per_token": 1.1e-06, + "input_cost_per_token_batches": 5.5e-07, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5.5e-06, + "output_cost_per_token_batches": 2.75e-06, + "prompt_cache_min_tokens": 4096, "source": "https://aws.amazon.com/about-aws/whats-new/2025/10/claude-4-5-haiku-anthropic-amazon-bedrock", "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096, - "input_cost_per_token_batches": 5.5e-07, - "output_cost_per_token_batches": 2.75e-06 + "supports_tool_search": true, + "supports_vision": true }, "eu.anthropic.claude-3-5-sonnet-20240620-v1:0": { + "cache_creation_input_token_cost": 3.75e-06, + "cache_read_input_token_cost": 3e-07, "input_cost_per_token": 3e-06, "litellm_provider": "bedrock", "max_input_tokens": 200000, @@ -17430,11 +17450,11 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-07, - "cache_creation_input_token_cost": 3.75e-06 + "supports_vision": true }, "eu.anthropic.claude-3-5-sonnet-20241022-v2:0": { + "cache_creation_input_token_cost": 3.75e-06, + "cache_read_input_token_cost": 3e-07, "input_cost_per_token": 3e-06, "litellm_provider": "bedrock", "max_input_tokens": 200000, @@ -17449,11 +17469,11 @@ "supports_prompt_caching": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-07, - "cache_creation_input_token_cost": 3.75e-06 + "supports_vision": true }, "eu.anthropic.claude-3-7-sonnet-20250219-v1:0": { + "cache_creation_input_token_cost": 3.75e-06, + "cache_read_input_token_cost": 3e-07, "input_cost_per_token": 3e-06, "litellm_provider": "bedrock", "max_input_tokens": 200000, @@ -17469,11 +17489,11 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-07, - "cache_creation_input_token_cost": 3.75e-06 + "supports_vision": true }, "eu.anthropic.claude-3-haiku-20240307-v1:0": { + "cache_creation_input_token_cost": 3.125e-07, + "cache_read_input_token_cost": 2.5e-08, "deprecation_date": "2026-09-10", "input_cost_per_token": 2.5e-07, "litellm_provider": "bedrock", @@ -17486,11 +17506,11 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 2.5e-08, - "cache_creation_input_token_cost": 3.125e-07 + "supports_vision": true }, "eu.anthropic.claude-3-opus-20240229-v1:0": { + "cache_creation_input_token_cost": 1.875e-05, + "cache_read_input_token_cost": 1.5e-06, "input_cost_per_token": 1.5e-05, "litellm_provider": "bedrock", "max_input_tokens": 200000, @@ -17501,11 +17521,11 @@ "supports_function_calling": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 1.5e-06, - "cache_creation_input_token_cost": 1.875e-05 + "supports_vision": true }, "eu.anthropic.claude-3-sonnet-20240229-v1:0": { + "cache_creation_input_token_cost": 3.75e-06, + "cache_read_input_token_cost": 3e-07, "deprecation_date": "2026-07-30", "input_cost_per_token": 3e-06, "litellm_provider": "bedrock", @@ -17518,9 +17538,7 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-07, - "cache_creation_input_token_cost": 3.75e-06 + "supports_vision": true }, "eu.anthropic.claude-opus-4-1-20250805-v1:0": { "cache_creation_input_token_cost": 1.875e-05, @@ -17532,6 +17550,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -17545,8 +17564,7 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "eu.anthropic.claude-opus-4-20250514-v1:0": { "cache_creation_input_token_cost": 1.875e-05, @@ -17558,6 +17576,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -17571,24 +17590,25 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "eu.anthropic.claude-sonnet-4-20250514-v1:0": { + "bedrock_converse_supports_strict_tools": false, "cache_creation_input_token_cost": 3.75e-06, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, "deprecation_date": "2026-10-14", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -17602,27 +17622,27 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "bedrock_converse_supports_strict_tools": false, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "eu.anthropic.claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 4.125e-06, "cache_creation_input_token_cost_above_1hr": 6.6e-06, + "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.32e-05, + "cache_creation_input_token_cost_above_200k_tokens": 8.25e-06, "cache_read_input_token_cost": 3.3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6.6e-07, "input_cost_per_token": 3.3e-06, "input_cost_per_token_above_200k_tokens": 6.6e-06, - "output_cost_per_token_above_200k_tokens": 2.475e-05, - "cache_creation_input_token_cost_above_200k_tokens": 8.25e-06, - "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.32e-05, - "cache_read_input_token_cost_above_200k_tokens": 6.6e-07, + "input_cost_per_token_batches": 1.65e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.65e-05, + "output_cost_per_token_above_200k_tokens": 2.475e-05, + "output_cost_per_token_batches": 8.25e-06, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -17631,17 +17651,15 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024, - "input_cost_per_token_batches": 1.65e-06, - "output_cost_per_token_batches": 8.25e-06 + "supports_tool_search": true, + "supports_vision": true }, "eu.meta.llama3-2-1b-instruct-v1:0": { "input_cost_per_token": 1.3e-07, @@ -18618,12 +18636,12 @@ }, "fireworks_ai/accounts/fireworks/models/glm-4p6": { "input_cost_per_token": 5.5e-07, - "output_cost_per_token": 2.19e-06, "litellm_provider": "fireworks_ai", "max_input_tokens": 202800, "max_output_tokens": 202800, "max_tokens": 202800, "mode": "chat", + "output_cost_per_token": 2.19e-06, "source": "https://fireworks.ai/pricing", "supports_function_calling": true, "supports_reasoning": true, @@ -19542,6 +19560,11 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 4e-07, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/pricing#2_0flash", "supported_modalities": [ "text", @@ -19563,12 +19586,7 @@ "supports_tool_choice": true, "supports_url_context": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "supports_web_search": true }, "gemini-2.0-flash-001": { "cache_read_input_token_cost": 3.75e-08, @@ -19581,6 +19599,11 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 6e-07, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_modalities": [ "text", @@ -19600,12 +19623,7 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "supports_web_search": true }, "gemini-2.0-flash-lite": { "cache_read_input_token_cost": 1.875e-08, @@ -19617,6 +19635,11 @@ "max_output_tokens": 8192, "mode": "chat", "output_cost_per_token": 3e-07, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/models#gemini-2.0-flash", "supported_modalities": [ "text", @@ -19635,12 +19658,7 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "supports_web_search": true }, "gemini-2.0-flash-lite-001": { "cache_read_input_token_cost": 1.875e-08, @@ -19652,6 +19670,11 @@ "max_output_tokens": 8192, "mode": "chat", "output_cost_per_token": 3e-07, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/models#gemini-2.0-flash", "supported_modalities": [ "text", @@ -19670,16 +19693,11 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "supports_web_search": true }, "gemini-2.5-flash": { - "deprecation_date": "2026-10-20", "cache_read_input_token_cost": 3e-08, + "deprecation_date": "2026-10-20", "input_cost_per_audio_token": 1e-06, "input_cost_per_token": 3e-07, "litellm_provider": "vertex_ai-language-models", @@ -19689,6 +19707,11 @@ "mode": "chat", "output_cost_per_reasoning_token": 2.5e-06, "output_cost_per_token": 2.5e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-preview", "supported_endpoints": [ "/v1/chat/completions", @@ -19706,6 +19729,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -19715,17 +19739,11 @@ "supports_tool_choice": true, "supports_url_context": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "supports_image_size": false + "supports_web_search": true }, "gemini-2.5-flash-image": { - "deprecation_date": "2026-10-02", "cache_read_input_token_cost": 3e-08, + "deprecation_date": "2026-10-02", "input_cost_per_audio_token": 1e-06, "input_cost_per_token": 3e-07, "litellm_provider": "vertex_ai-language-models", @@ -19756,6 +19774,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -19765,8 +19784,7 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": false, - "tpm": 8000000, - "supports_image_size": false + "tpm": 8000000 }, "gemini-3-pro-image": { "deprecation_date": "2027-05-28", @@ -19782,6 +19800,11 @@ "output_cost_per_image_token": 0.00012, "output_cost_per_token": 1.2e-05, "output_cost_per_token_batches": 6e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-3-pro-image", "supported_endpoints": [ "/v1/chat/completions", @@ -19799,16 +19822,11 @@ "supports_function_calling": false, "supports_prompt_caching": true, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_vision": true, "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, - "web_search_billing_unit": "per_query", - "supports_service_tier": true + "web_search_billing_unit": "per_query" }, "gemini-3-pro-image-preview": { "input_cost_per_image": 0.0011, @@ -19823,6 +19841,11 @@ "output_cost_per_image_token": 0.00012, "output_cost_per_token": 1.2e-05, "output_cost_per_token_batches": 6e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -19843,11 +19866,6 @@ "supports_system_messages": true, "supports_vision": true, "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini-3.1-flash-image": { @@ -19862,6 +19880,11 @@ "output_cost_per_image": 0.0672, "output_cost_per_image_token": 6e-05, "output_cost_per_token": 3e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models", "supported_endpoints": [ "/v1/chat/completions", @@ -19882,11 +19905,6 @@ "supports_system_messages": true, "supports_vision": true, "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini-3.1-flash-image-preview": { @@ -19900,6 +19918,11 @@ "output_cost_per_image": 0.0672, "output_cost_per_image_token": 6e-05, "output_cost_per_token": 3e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models", "supported_endpoints": [ "/v1/chat/completions", @@ -19920,11 +19943,6 @@ "supports_system_messages": true, "supports_vision": true, "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini-3.1-flash-lite-image": { @@ -19967,6 +19985,7 @@ }, "gemini-3.1-flash-lite-preview": { "cache_read_input_token_cost": 2.5e-08, + "deprecation_date": "2026-05-25", "input_cost_per_audio_token": 5e-07, "input_cost_per_token": 2.5e-07, "litellm_provider": "vertex_ai-language-models", @@ -19976,6 +19995,11 @@ "mode": "chat", "output_cost_per_reasoning_token": 1.5e-06, "output_cost_per_token": 1.5e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/models", "supported_endpoints": [ "/v1/chat/completions", @@ -19994,6 +20018,7 @@ "supports_audio_input": true, "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -20005,35 +20030,38 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, - "web_search_billing_unit": "per_query", - "deprecation_date": "2026-05-25" + "web_search_billing_unit": "per_query" }, "gemini-3.1-flash-lite": { - "deprecation_date": "2027-05-07", "cache_read_input_token_cost": 2.5e-08, "cache_read_input_token_cost_flex": 1.25e-08, + "cache_read_input_token_cost_per_audio_token": 5e-08, "cache_read_input_token_cost_priority": 4.5e-08, + "deprecation_date": "2027-05-07", "input_cost_per_audio_token": 5e-07, "input_cost_per_token": 2.5e-07, "input_cost_per_token_batches": 1.25e-07, "input_cost_per_token_flex": 1.25e-07, "input_cost_per_token_priority": 4.5e-07, "litellm_provider": "vertex_ai-language-models", + "max_audio_length_hours": 8.4, + "max_audio_per_prompt": 1, + "max_images_per_prompt": 3000, "max_input_tokens": 1048576, "max_output_tokens": 65536, "max_tokens": 65536, + "max_videos_per_prompt": 10, "mode": "chat", "output_cost_per_reasoning_token": 1.5e-06, "output_cost_per_token": 1.5e-06, "output_cost_per_token_batches": 7.5e-07, "output_cost_per_token_flex": 7.5e-07, "output_cost_per_token_priority": 2.7e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-3.1-flash-lite", "supported_endpoints": [ "/v1/chat/completions", @@ -20051,39 +20079,29 @@ ], "supports_audio_input": true, "supports_audio_output": false, + "supports_code_execution": true, + "supports_file_search": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_tool_choice": true, "supports_url_context": true, "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, - "web_search_billing_unit": "per_query", - "cache_read_input_token_cost_per_audio_token": 5e-08, - "max_audio_length_hours": 8.4, - "max_audio_per_prompt": 1, - "max_images_per_prompt": 3000, - "max_videos_per_prompt": 10, - "supports_code_execution": true, - "supports_file_search": true, - "supports_service_tier": true + "web_search_billing_unit": "per_query" }, "gemini-3.5-flash-lite": { - "deprecation_date": "2027-07-21", "cache_read_input_token_cost": 3e-08, "cache_read_input_token_cost_flex": 2e-08, "cache_read_input_token_cost_priority": 5e-08, + "deprecation_date": "2027-07-21", "input_cost_per_token": 3e-07, "input_cost_per_token_batches": 1.5e-07, "input_cost_per_token_flex": 1.5e-07, @@ -20098,6 +20116,11 @@ "output_cost_per_token_batches": 1.25e-06, "output_cost_per_token_flex": 1.25e-06, "output_cost_per_token_priority": 4.5e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -20116,6 +20139,7 @@ "supports_audio_input": true, "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -20127,12 +20151,6 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "deep-research-pro-preview-12-2025": { @@ -20170,8 +20188,8 @@ "supports_web_search": true }, "gemini-2.5-flash-lite": { - "deprecation_date": "2026-10-20", "cache_read_input_token_cost": 1e-08, + "deprecation_date": "2026-10-20", "input_cost_per_audio_token": 3e-07, "input_cost_per_token": 1e-07, "litellm_provider": "vertex_ai-language-models", @@ -20181,6 +20199,11 @@ "mode": "chat", "output_cost_per_reasoning_token": 4e-07, "output_cost_per_token": 4e-07, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-preview", "supported_endpoints": [ "/v1/chat/completions", @@ -20198,6 +20221,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -20207,13 +20231,7 @@ "supports_tool_choice": true, "supports_url_context": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "supports_image_size": false + "supports_web_search": true }, "gemini-2.5-flash-lite-preview-09-2025": { "cache_read_input_token_cost": 1e-08, @@ -20226,6 +20244,11 @@ "mode": "chat", "output_cost_per_reasoning_token": 4e-07, "output_cost_per_token": 4e-07, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://developers.googleblog.com/en/continuing-to-bring-you-our-latest-models-with-an-improved-gemini-2-5-flash-and-flash-lite-release/", "supported_endpoints": [ "/v1/chat/completions", @@ -20243,6 +20266,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -20252,13 +20276,7 @@ "supports_tool_choice": true, "supports_url_context": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "supports_image_size": false + "supports_web_search": true }, "gemini-2.5-flash-preview-09-2025": { "cache_read_input_token_cost": 7.5e-08, @@ -20271,6 +20289,11 @@ "mode": "chat", "output_cost_per_reasoning_token": 2.5e-06, "output_cost_per_token": 2.5e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://developers.googleblog.com/en/continuing-to-bring-you-our-latest-models-with-an-improved-gemini-2-5-flash-and-flash-lite-release/", "supported_endpoints": [ "/v1/chat/completions", @@ -20288,6 +20311,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -20297,16 +20321,11 @@ "supports_tool_choice": true, "supports_url_context": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "supports_image_size": false + "supports_web_search": true }, "gemini-live-2.5-flash-preview-native-audio-09-2025": { "cache_read_input_token_cost": 7.5e-08, + "gemini_native_audio": true, "input_cost_per_audio_token": 3e-06, "input_cost_per_token": 3e-07, "litellm_provider": "vertex_ai-language-models", @@ -20316,6 +20335,11 @@ "mode": "realtime", "output_cost_per_audio_token": 1.2e-05, "output_cost_per_token": 2e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_endpoints": [ "/vertex_ai/live" @@ -20341,16 +20365,11 @@ "supports_tool_choice": true, "supports_url_context": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "gemini_native_audio": true + "supports_web_search": true }, "gemini/gemini-live-2.5-flash-preview-native-audio-09-2025": { "cache_read_input_token_cost": 7.5e-08, + "gemini_native_audio": true, "input_cost_per_audio_token": 3e-06, "input_cost_per_token": 3e-07, "litellm_provider": "gemini", @@ -20361,6 +20380,11 @@ "output_cost_per_audio_token": 1.2e-05, "output_cost_per_token": 2e-06, "rpm": 100000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_endpoints": [ "/v1/realtime" @@ -20387,17 +20411,11 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 8000000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "gemini_native_audio": true + "tpm": 8000000 }, "gemini-2.5-flash-lite-preview-06-17": { - "deprecation_date": "2025-11-18", "cache_read_input_token_cost": 2.5e-08, + "deprecation_date": "2025-11-18", "input_cost_per_audio_token": 5e-07, "input_cost_per_token": 1e-07, "litellm_provider": "vertex_ai-language-models", @@ -20407,6 +20425,11 @@ "mode": "chat", "output_cost_per_reasoning_token": 4e-07, "output_cost_per_token": 4e-07, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-preview", "supported_endpoints": [ "/v1/chat/completions", @@ -20424,6 +20447,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -20433,19 +20457,13 @@ "supports_tool_choice": true, "supports_url_context": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "supports_image_size": false + "supports_web_search": true }, "gemini-2.5-pro": { - "deprecation_date": "2026-10-20", + "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "cache_read_input_token_cost": 1.25e-07, "cache_read_input_token_cost_above_200k_tokens": 2.5e-07, - "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, + "deprecation_date": "2026-10-20", "input_cost_per_token": 1.25e-06, "input_cost_per_token_above_200k_tokens": 2.5e-06, "litellm_provider": "vertex_ai-language-models", @@ -20455,6 +20473,11 @@ "mode": "chat", "output_cost_per_token": 1e-05, "output_cost_per_token_above_200k_tokens": 1.5e-05, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -20479,21 +20502,20 @@ "supports_tool_choice": true, "supports_video_input": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "supports_web_search": true }, "gemini-3-pro-preview": { - "deprecation_date": "2026-03-26", + "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, + "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, + "cache_read_input_token_cost_priority": 3.6e-07, + "deprecation_date": "2026-03-26", "input_cost_per_token": 2e-06, "input_cost_per_token_above_200k_tokens": 4e-06, + "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, "input_cost_per_token_batches": 1e-06, + "input_cost_per_token_priority": 3.6e-06, "litellm_provider": "vertex_ai-language-models", "max_input_tokens": 1048576, "max_output_tokens": 65535, @@ -20501,7 +20523,14 @@ "mode": "chat", "output_cost_per_token": 1.2e-05, "output_cost_per_token_above_200k_tokens": 1.8e-05, + "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, "output_cost_per_token_batches": 6e-06, + "output_cost_per_token_priority": 2.16e-05, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -20519,6 +20548,7 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -20528,37 +20558,36 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 3.6e-06, - "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, - "output_cost_per_token_priority": 2.16e-05, - "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, - "cache_read_input_token_cost_priority": 3.6e-07, - "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini-3.1-pro-preview": { - "prompt_cache_min_tokens": 4096, + "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, + "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, + "cache_read_input_token_cost_priority": 3.6e-07, "input_cost_per_token": 2e-06, "input_cost_per_token_above_200k_tokens": 4e-06, + "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, "input_cost_per_token_batches": 1e-06, + "input_cost_per_token_priority": 3.6e-06, "litellm_provider": "vertex_ai-language-models", "max_input_tokens": 1048576, "max_output_tokens": 65536, "max_tokens": 65536, "mode": "chat", + "output_cost_per_image": 0.00012, "output_cost_per_token": 1.2e-05, "output_cost_per_token_above_200k_tokens": 1.8e-05, + "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, "output_cost_per_token_batches": 6e-06, - "output_cost_per_image": 0.00012, + "output_cost_per_token_priority": 2.16e-05, + "prompt_cache_min_tokens": 4096, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models", "supported_endpoints": [ "/v1/chat/completions", @@ -20576,35 +20605,23 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, + "supports_url_context": true, "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_url_context": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 3.6e-06, - "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, - "output_cost_per_token_priority": 2.16e-05, - "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, - "cache_read_input_token_cost_priority": 3.6e-07, - "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini-3.1-pro-preview-customtools": { - "prompt_cache_min_tokens": 4096, + "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "input_cost_per_token": 2e-06, "input_cost_per_token_above_200k_tokens": 4e-06, "input_cost_per_token_batches": 1e-06, @@ -20613,10 +20630,16 @@ "max_output_tokens": 65536, "max_tokens": 65536, "mode": "chat", + "output_cost_per_image": 0.00012, "output_cost_per_token": 1.2e-05, "output_cost_per_token_above_200k_tokens": 1.8e-05, "output_cost_per_token_batches": 6e-06, - "output_cost_per_image": 0.00012, + "prompt_cache_min_tokens": 4096, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models", "supported_endpoints": [ "/v1/chat/completions", @@ -20634,31 +20657,30 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, + "supports_url_context": true, "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_url_context": true, - "supports_native_streaming": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "vertex_ai/gemini-3-pro-preview": { + "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, + "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, + "cache_read_input_token_cost_priority": 3.6e-07, "input_cost_per_token": 2e-06, "input_cost_per_token_above_200k_tokens": 4e-06, + "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, "input_cost_per_token_batches": 1e-06, + "input_cost_per_token_priority": 3.6e-06, "litellm_provider": "vertex_ai", "max_input_tokens": 1048576, "max_output_tokens": 65535, @@ -20666,7 +20688,14 @@ "mode": "chat", "output_cost_per_token": 1.2e-05, "output_cost_per_token_above_200k_tokens": 1.8e-05, + "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, "output_cost_per_token_batches": 6e-06, + "output_cost_per_token_priority": 2.16e-05, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -20684,6 +20713,7 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -20693,30 +20723,27 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 3.6e-06, - "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, - "output_cost_per_token_priority": 2.16e-05, - "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, - "cache_read_input_token_cost_priority": 3.6e-07, - "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "vertex_ai/gemini-3-flash-preview": { "cache_read_input_token_cost": 5e-08, - "input_cost_per_token": 5e-07, + "cache_read_input_token_cost_priority": 9e-08, "input_cost_per_audio_token": 1e-06, + "input_cost_per_audio_token_priority": 1.8e-06, + "input_cost_per_token": 5e-07, + "input_cost_per_token_priority": 9e-07, "litellm_provider": "vertex_ai", "max_input_tokens": 1048576, "max_output_tokens": 65535, "max_tokens": 65535, "mode": "chat", "output_cost_per_token": 3e-06, + "output_cost_per_token_priority": 5.4e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -20734,6 +20761,7 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -20743,24 +20771,19 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 9e-07, - "input_cost_per_audio_token_priority": 1.8e-06, - "output_cost_per_token_priority": 5.4e-06, - "cache_read_input_token_cost_priority": 9e-08, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "vertex_ai/gemini-3.5-flash": { - "prompt_cache_min_tokens": 4096, - "deprecation_date": "2027-05-19", "cache_read_input_token_cost": 1.5e-07, - "input_cost_per_token": 1.5e-06, + "cache_read_input_token_cost_flex": 7.5e-08, + "cache_read_input_token_cost_priority": 2.7e-07, + "deprecation_date": "2027-05-19", "input_cost_per_audio_token": 1.5e-06, + "input_cost_per_audio_token_priority": 2.7e-06, + "input_cost_per_token": 1.5e-06, + "input_cost_per_token_batches": 7.5e-07, + "input_cost_per_token_flex": 7.5e-07, + "input_cost_per_token_priority": 2.7e-06, "litellm_provider": "vertex_ai", "max_input_tokens": 1048576, "max_output_tokens": 65535, @@ -20768,7 +20791,16 @@ "mode": "chat", "output_cost_per_reasoning_token": 9e-06, "output_cost_per_token": 9e-06, + "output_cost_per_token_batches": 4.5e-06, + "output_cost_per_token_flex": 4.5e-06, + "output_cost_per_token_priority": 1.62e-05, + "prompt_cache_min_tokens": 4096, "regional_endpoint_uplift_multiplier": 1.1, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -20786,6 +20818,7 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -20797,30 +20830,16 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 2.7e-06, - "input_cost_per_audio_token_priority": 2.7e-06, - "output_cost_per_token_priority": 1.62e-05, - "cache_read_input_token_cost_priority": 2.7e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, - "web_search_billing_unit": "per_query", - "input_cost_per_token_batches": 7.5e-07, - "output_cost_per_token_batches": 4.5e-06, - "input_cost_per_token_flex": 7.5e-07, - "output_cost_per_token_flex": 4.5e-06, - "cache_read_input_token_cost_flex": 7.5e-08 + "web_search_billing_unit": "per_query" }, "vertex_ai/gemini-3.6-flash": { - "prompt_cache_min_tokens": 4096, "cache_read_input_token_cost": 7.5e-08, "cache_read_input_token_cost_flex": 3.75e-08, + "cache_read_input_token_cost_priority": 1.35e-07, "input_cost_per_token": 7.5e-07, "input_cost_per_token_batches": 3.75e-07, "input_cost_per_token_flex": 3.75e-07, + "input_cost_per_token_priority": 1.35e-06, "litellm_provider": "vertex_ai", "max_input_tokens": 1048576, "max_output_tokens": 65536, @@ -20830,7 +20849,14 @@ "output_cost_per_token": 3.75e-06, "output_cost_per_token_batches": 1.875e-06, "output_cost_per_token_flex": 1.875e-06, + "output_cost_per_token_priority": 6.75e-06, + "prompt_cache_min_tokens": 4096, "regional_endpoint_uplift_multiplier": 1.1, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -20848,6 +20874,7 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -20859,24 +20886,16 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 1.35e-06, - "output_cost_per_token_priority": 6.75e-06, - "cache_read_input_token_cost_priority": 1.35e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "vertex_ai/gemini-3.7-flash": { - "prompt_cache_min_tokens": 4096, "cache_read_input_token_cost": 7.5e-08, "cache_read_input_token_cost_flex": 3.75e-08, + "cache_read_input_token_cost_priority": 1.35e-07, "input_cost_per_token": 7.5e-07, "input_cost_per_token_batches": 3.75e-07, "input_cost_per_token_flex": 3.75e-07, + "input_cost_per_token_priority": 1.35e-06, "litellm_provider": "vertex_ai", "max_input_tokens": 1048576, "max_output_tokens": 65536, @@ -20886,7 +20905,14 @@ "output_cost_per_token": 3.75e-06, "output_cost_per_token_batches": 1.875e-06, "output_cost_per_token_flex": 1.875e-06, + "output_cost_per_token_priority": 6.75e-06, + "prompt_cache_min_tokens": 4096, "regional_endpoint_uplift_multiplier": 1.1, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -20904,6 +20930,7 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -20915,34 +20942,36 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 1.35e-06, - "output_cost_per_token_priority": 6.75e-06, - "cache_read_input_token_cost_priority": 1.35e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "vertex_ai/gemini-3.1-pro-preview": { - "prompt_cache_min_tokens": 4096, + "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, + "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, + "cache_read_input_token_cost_priority": 3.6e-07, "input_cost_per_token": 2e-06, "input_cost_per_token_above_200k_tokens": 4e-06, + "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, "input_cost_per_token_batches": 1e-06, + "input_cost_per_token_priority": 3.6e-06, "litellm_provider": "vertex_ai", "max_input_tokens": 1048576, "max_output_tokens": 65536, "max_tokens": 65536, "mode": "chat", + "output_cost_per_image": 0.00012, "output_cost_per_token": 1.2e-05, "output_cost_per_token_above_200k_tokens": 1.8e-05, + "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, "output_cost_per_token_batches": 6e-06, - "output_cost_per_image": 0.00012, + "output_cost_per_token_priority": 2.16e-05, + "prompt_cache_min_tokens": 4096, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models", "supported_endpoints": [ "/v1/chat/completions", @@ -20960,47 +20989,47 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, + "supports_url_context": true, "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_url_context": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 3.6e-06, - "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, - "output_cost_per_token_priority": 2.16e-05, - "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, - "cache_read_input_token_cost_priority": 3.6e-07, - "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "vertex_ai/gemini-3.1-pro-preview-customtools": { - "prompt_cache_min_tokens": 4096, + "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, + "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, + "cache_read_input_token_cost_priority": 3.6e-07, "input_cost_per_token": 2e-06, "input_cost_per_token_above_200k_tokens": 4e-06, + "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, "input_cost_per_token_batches": 1e-06, + "input_cost_per_token_priority": 3.6e-06, "litellm_provider": "vertex_ai", "max_input_tokens": 1048576, "max_output_tokens": 65536, "max_tokens": 65536, "mode": "chat", + "output_cost_per_image": 0.00012, "output_cost_per_token": 1.2e-05, "output_cost_per_token_above_200k_tokens": 1.8e-05, + "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, "output_cost_per_token_batches": 6e-06, - "output_cost_per_image": 0.00012, + "output_cost_per_token_priority": 2.16e-05, + "prompt_cache_min_tokens": 4096, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models", "supported_endpoints": [ "/v1/chat/completions", @@ -21018,28 +21047,17 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, + "supports_url_context": true, "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_url_context": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 3.6e-06, - "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, - "output_cost_per_token_priority": 2.16e-05, - "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, - "cache_read_input_token_cost_priority": 3.6e-07, - "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini-2.5-pro-preview-tts": { @@ -21055,6 +21073,11 @@ "mode": "chat", "output_cost_per_token": 1e-05, "output_cost_per_token_above_200k_tokens": 1.5e-05, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-2.5-pro-preview", "supported_modalities": [ "text" @@ -21070,24 +21093,19 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "supports_web_search": true }, "gemini-robotics-er-1.5-preview": { "cache_read_input_token_cost": 0, - "input_cost_per_token": 3e-07, "input_cost_per_audio_token": 1e-06, + "input_cost_per_token": 3e-07, "litellm_provider": "vertex_ai-language-models", "max_input_tokens": 1048576, "max_output_tokens": 65535, "max_tokens": 65535, "mode": "chat", - "output_cost_per_token": 2.5e-06, "output_cost_per_reasoning_token": 2.5e-06, + "output_cost_per_token": 2.5e-06, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-robotics-er-1-5-preview", "supported_endpoints": [ "/v1/chat/completions", @@ -21116,15 +21134,21 @@ "gemini/gemini-robotics-er-1.5-preview": { "cache_read_input_token_cost": 0, "deprecation_date": "2026-04-30", - "input_cost_per_token": 3e-07, "input_cost_per_audio_token": 1e-06, + "input_cost_per_token": 3e-07, "litellm_provider": "gemini", "max_input_tokens": 1048576, "max_output_tokens": 65535, "max_tokens": 65535, "mode": "chat", - "output_cost_per_token": 2.5e-06, "output_cost_per_reasoning_token": 2.5e-06, + "output_cost_per_token": 2.5e-06, + "rpm": 10, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-robotics-er-1-5-preview", "supported_endpoints": [ "/v1/chat/completions", @@ -21150,13 +21174,7 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 250000, - "rpm": 10, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 250000 }, "gemini/gemini-robotics-er-2-preview": { "cache_read_input_token_cost": 2e-07, @@ -21432,6 +21450,11 @@ "mode": "chat", "output_cost_per_token": 4e-07, "rpm": 10000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/pricing#2_0flash", "supported_modalities": [ "text", @@ -21453,12 +21476,7 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 10000000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 10000000 }, "gemini/gemini-2.0-flash-001": { "cache_read_input_token_cost": 2.5e-08, @@ -21472,6 +21490,11 @@ "mode": "chat", "output_cost_per_token": 4e-07, "rpm": 10000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/pricing#2_0flash", "supported_modalities": [ "text", @@ -21491,12 +21514,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "tpm": 10000000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 10000000 }, "gemini/gemini-2.0-flash-lite": { "cache_read_input_token_cost": 1.875e-08, @@ -21509,6 +21527,11 @@ "mode": "chat", "output_cost_per_token": 3e-07, "rpm": 4000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-2.0-flash-lite", "supported_modalities": [ "text", @@ -21527,12 +21550,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "tpm": 4000000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 4000000 }, "gemini/gemini-2.5-flash": { "cache_read_input_token_cost": 3e-08, @@ -21546,6 +21564,11 @@ "output_cost_per_reasoning_token": 2.5e-06, "output_cost_per_token": 2.5e-06, "rpm": 100000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-preview", "supported_endpoints": [ "/v1/chat/completions", @@ -21563,6 +21586,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -21573,13 +21597,7 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 8000000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "supports_image_size": false + "tpm": 8000000 }, "gemini/gemini-2.5-flash-image": { "cache_read_input_token_cost": 3e-08, @@ -21587,7 +21605,6 @@ "input_cost_per_audio_token": 1e-06, "input_cost_per_token": 3e-07, "litellm_provider": "gemini", - "supports_reasoning": false, "max_input_tokens": 32768, "max_output_tokens": 32768, "max_tokens": 32768, @@ -21597,6 +21614,11 @@ "output_cost_per_reasoning_token": 2.5e-06, "output_cost_per_token": 2.5e-06, "rpm": 100000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-2.5-flash-image", "supported_endpoints": [ "/v1/chat/completions", @@ -21615,22 +21637,18 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, + "supports_reasoning": false, "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 8000000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "supports_image_size": false + "tpm": 8000000 }, "gemini/gemini-3-pro-image": { "input_cost_per_image": 0.0011, @@ -21644,9 +21662,13 @@ "output_cost_per_image": 0.134, "output_cost_per_image_token": 0.00012, "output_cost_per_token": 1.2e-05, - "rpm": 1000, - "tpm": 4000000, "output_cost_per_token_batches": 6e-06, + "rpm": 1000, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-3-pro-image", "supported_endpoints": [ "/v1/chat/completions", @@ -21663,18 +21685,14 @@ ], "supports_function_calling": false, "supports_prompt_caching": true, + "supports_reasoning": false, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_vision": true, "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, - "web_search_billing_unit": "per_query", - "supports_reasoning": false, - "supports_service_tier": true + "tpm": 4000000, + "web_search_billing_unit": "per_query" }, "gemini/gemini-3-pro-image-preview": { "deprecation_date": "2026-06-25", @@ -21689,9 +21707,13 @@ "output_cost_per_image": 0.134, "output_cost_per_image_token": 0.00012, "output_cost_per_token": 1.2e-05, - "rpm": 1000, - "tpm": 4000000, "output_cost_per_token_batches": 6e-06, + "rpm": 1000, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -21713,11 +21735,7 @@ "supports_system_messages": true, "supports_vision": true, "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, + "tpm": 4000000, "web_search_billing_unit": "per_query" }, "gemini/gemini-3.1-flash-image": { @@ -21730,10 +21748,15 @@ "mode": "image_generation", "output_cost_per_image": 0.045, "output_cost_per_image_token": 6e-05, + "output_cost_per_image_token_batches": 3e-05, "output_cost_per_token": 3e-06, "output_cost_per_token_batches": 1.5e-06, "rpm": 1000, - "tpm": 4000000, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-3.1-flash-image", "supported_endpoints": [ "/v1/chat/completions", @@ -21755,13 +21778,8 @@ "supports_system_messages": true, "supports_vision": true, "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, - "web_search_billing_unit": "per_query", - "output_cost_per_image_token_batches": 3e-05 + "tpm": 4000000, + "web_search_billing_unit": "per_query" }, "gemini/gemini-3.1-flash-image-preview": { "deprecation_date": "2026-06-25", @@ -21777,7 +21795,11 @@ "output_cost_per_token": 3e-06, "output_cost_per_token_batches": 1.5e-06, "rpm": 1000, - "tpm": 4000000, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-3.1-flash-image-preview", "supported_endpoints": [ "/v1/chat/completions", @@ -21799,11 +21821,7 @@ "supports_system_messages": true, "supports_vision": true, "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, + "tpm": 4000000, "web_search_billing_unit": "per_query" }, "gemini/gemini-3.1-flash-lite-image": { @@ -21854,9 +21872,13 @@ "output_cost_per_image": 0.134, "output_cost_per_image_token": 0.00012, "output_cost_per_token": 1.2e-05, - "rpm": 1000, - "tpm": 4000000, "output_cost_per_token_batches": 6e-06, + "rpm": 1000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -21877,11 +21899,7 @@ "supports_system_messages": true, "supports_vision": true, "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 4000000 }, "gemini/gemini-2.5-flash-lite": { "cache_read_input_token_cost": 1e-08, @@ -21895,6 +21913,11 @@ "output_cost_per_reasoning_token": 4e-07, "output_cost_per_token": 4e-07, "rpm": 15, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-lite", "supported_endpoints": [ "/v1/chat/completions", @@ -21912,6 +21935,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -21922,13 +21946,7 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 250000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "supports_image_size": false + "tpm": 250000 }, "gemini/gemini-2.5-flash-lite-preview-09-2025": { "cache_read_input_token_cost": 1e-08, @@ -21943,6 +21961,11 @@ "output_cost_per_reasoning_token": 4e-07, "output_cost_per_token": 4e-07, "rpm": 15, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://developers.googleblog.com/en/continuing-to-bring-you-our-latest-models-with-an-improved-gemini-2-5-flash-and-flash-lite-release/", "supported_endpoints": [ "/v1/chat/completions", @@ -21960,6 +21983,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -21970,13 +21994,7 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 250000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "supports_image_size": false + "tpm": 250000 }, "gemini/gemini-2.5-flash-preview-09-2025": { "cache_read_input_token_cost": 7.5e-08, @@ -21991,6 +22009,11 @@ "output_cost_per_reasoning_token": 2.5e-06, "output_cost_per_token": 2.5e-06, "rpm": 15, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://developers.googleblog.com/en/continuing-to-bring-you-our-latest-models-with-an-improved-gemini-2-5-flash-and-flash-lite-release/", "supported_endpoints": [ "/v1/chat/completions", @@ -22008,6 +22031,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -22018,13 +22042,7 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 250000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "supports_image_size": false + "tpm": 250000 }, "gemini/gemini-flash-latest": { "cache_read_input_token_cost": 7.5e-08, @@ -22038,6 +22056,11 @@ "output_cost_per_reasoning_token": 2.5e-06, "output_cost_per_token": 2.5e-06, "rpm": 15, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://developers.googleblog.com/en/continuing-to-bring-you-our-latest-models-with-an-improved-gemini-2-5-flash-and-flash-lite-release/", "supported_endpoints": [ "/v1/chat/completions", @@ -22065,12 +22088,7 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 250000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 250000 }, "gemini/gemini-flash-lite-latest": { "cache_read_input_token_cost": 2.5e-08, @@ -22084,6 +22102,11 @@ "output_cost_per_reasoning_token": 4e-07, "output_cost_per_token": 4e-07, "rpm": 15, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://developers.googleblog.com/en/continuing-to-bring-you-our-latest-models-with-an-improved-gemini-2-5-flash-and-flash-lite-release/", "supported_endpoints": [ "/v1/chat/completions", @@ -22111,16 +22134,11 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 250000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 250000 }, "gemini/gemini-2.5-flash-lite-preview-06-17": { - "deprecation_date": "2025-11-18", "cache_read_input_token_cost": 2.5e-08, + "deprecation_date": "2025-11-18", "input_cost_per_audio_token": 5e-07, "input_cost_per_token": 1e-07, "litellm_provider": "gemini", @@ -22131,6 +22149,11 @@ "output_cost_per_reasoning_token": 4e-07, "output_cost_per_token": 4e-07, "rpm": 15, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-lite", "supported_endpoints": [ "/v1/chat/completions", @@ -22148,6 +22171,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -22158,33 +22182,27 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 250000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - }, - "supports_image_size": false + "tpm": 250000 }, "gemini/gemini-2.5-flash-preview-tts": { "input_cost_per_token": 5e-07, "litellm_provider": "gemini", "mode": "audio_speech", "output_cost_per_token": 1e-05, + "rpm": 10, "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_endpoints": [ "/v1/audio/speech" ], - "tpm": 4000000, - "rpm": 10 + "tpm": 4000000 }, "gemini/gemini-2.5-pro": { "cache_read_input_token_cost": 1.25e-07, "cache_read_input_token_cost_above_200k_tokens": 2.5e-07, "input_cost_per_token": 1.25e-06, "input_cost_per_token_above_200k_tokens": 2.5e-06, - "input_cost_per_token_priority": 1.25e-06, "input_cost_per_token_above_200k_tokens_priority": 2.5e-06, + "input_cost_per_token_priority": 1.25e-06, "litellm_provider": "gemini", "max_input_tokens": 1048576, "max_output_tokens": 65535, @@ -22192,9 +22210,14 @@ "mode": "chat", "output_cost_per_token": 1e-05, "output_cost_per_token_above_200k_tokens": 1.5e-05, - "output_cost_per_token_priority": 1e-05, "output_cost_per_token_above_200k_tokens_priority": 1.5e-05, + "output_cost_per_token_priority": 1e-05, "rpm": 2000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -22220,12 +22243,7 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "tpm": 800000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 800000 }, "gemini/gemini-2.5-computer-use-preview-10-2025": { "input_cost_per_token": 1.25e-06, @@ -22258,12 +22276,16 @@ "tpm": 800000 }, "gemini/gemini-3-pro-preview": { - "deprecation_date": "2026-03-09", "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, + "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, + "cache_read_input_token_cost_priority": 3.6e-07, + "deprecation_date": "2026-03-09", "input_cost_per_token": 2e-06, "input_cost_per_token_above_200k_tokens": 4e-06, + "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, "input_cost_per_token_batches": 1e-06, + "input_cost_per_token_priority": 3.6e-06, "litellm_provider": "gemini", "max_input_tokens": 1048576, "max_output_tokens": 65535, @@ -22271,8 +22293,15 @@ "mode": "chat", "output_cost_per_token": 1.2e-05, "output_cost_per_token_above_200k_tokens": 1.8e-05, + "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, "output_cost_per_token_batches": 6e-06, + "output_cost_per_token_priority": 2.16e-05, "rpm": 2000, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -22300,17 +22329,6 @@ "supports_vision": true, "supports_web_search": true, "tpm": 800000, - "input_cost_per_token_priority": 3.6e-06, - "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, - "output_cost_per_token_priority": 2.16e-05, - "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, - "cache_read_input_token_cost_priority": 3.6e-07, - "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini/gemini-3.1-flash-lite-preview": { @@ -22326,6 +22344,11 @@ "output_cost_per_reasoning_token": 1.5e-06, "output_cost_per_token": 1.5e-06, "rpm": 15, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/models", "supported_endpoints": [ "/v1/chat/completions", @@ -22344,6 +22367,7 @@ "supports_audio_input": true, "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -22355,18 +22379,13 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, "tpm": 250000, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini/gemini-3.1-flash-lite": { "cache_read_input_token_cost": 2.5e-08, "cache_read_input_token_cost_flex": 1.25e-08, + "cache_read_input_token_cost_per_audio_token": 5e-08, "cache_read_input_token_cost_priority": 4.5e-08, "deprecation_date": "2027-05-07", "input_cost_per_audio_token": 5e-07, @@ -22375,9 +22394,13 @@ "input_cost_per_token_flex": 1.25e-07, "input_cost_per_token_priority": 4.5e-07, "litellm_provider": "gemini", + "max_audio_length_hours": 8.4, + "max_audio_per_prompt": 1, + "max_images_per_prompt": 3000, "max_input_tokens": 1048576, "max_output_tokens": 65536, "max_tokens": 65536, + "max_videos_per_prompt": 10, "mode": "chat", "output_cost_per_reasoning_token": 1.5e-06, "output_cost_per_token": 1.5e-06, @@ -22385,6 +22408,11 @@ "output_cost_per_token_flex": 7.5e-07, "output_cost_per_token_priority": 2.7e-06, "rpm": 15, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-3.1-flash-lite", "supported_endpoints": [ "/v1/chat/completions", @@ -22402,34 +22430,24 @@ ], "supports_audio_input": true, "supports_audio_output": false, + "supports_code_execution": true, + "supports_file_search": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_tool_choice": true, "supports_url_context": true, "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, "tpm": 250000, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, - "web_search_billing_unit": "per_query", - "cache_read_input_token_cost_per_audio_token": 5e-08, - "max_audio_length_hours": 8.4, - "max_audio_per_prompt": 1, - "max_images_per_prompt": 3000, - "max_videos_per_prompt": 10, - "supports_code_execution": true, - "supports_file_search": true, - "supports_service_tier": true + "web_search_billing_unit": "per_query" }, "gemini/gemini-3.5-flash-lite": { "cache_read_input_token_cost": 3e-08, @@ -22450,6 +22468,11 @@ "output_cost_per_token_flex": 1.25e-06, "output_cost_per_token_priority": 4.5e-06, "rpm": 15, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -22468,6 +22491,7 @@ "supports_audio_input": true, "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -22479,19 +22503,16 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, "tpm": 250000, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini/gemini-3-flash-preview": { "cache_read_input_token_cost": 5e-08, + "cache_read_input_token_cost_priority": 9e-08, "input_cost_per_audio_token": 1e-06, + "input_cost_per_audio_token_priority": 1.8e-06, "input_cost_per_token": 5e-07, + "input_cost_per_token_priority": 9e-07, "litellm_provider": "gemini", "max_input_tokens": 1048576, "max_output_tokens": 65535, @@ -22499,7 +22520,13 @@ "mode": "chat", "output_cost_per_reasoning_token": 3e-06, "output_cost_per_token": 3e-06, + "output_cost_per_token_priority": 5.4e-06, "rpm": 2000, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/pricing/gemini-3", "supported_endpoints": [ "/v1/chat/completions", @@ -22517,6 +22544,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -22527,24 +22555,19 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, "tpm": 800000, - "input_cost_per_token_priority": 9e-07, - "input_cost_per_audio_token_priority": 1.8e-06, - "output_cost_per_token_priority": 5.4e-06, - "cache_read_input_token_cost_priority": 9e-08, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini/gemini-3.5-flash": { - "prompt_cache_min_tokens": 4096, "cache_read_input_token_cost": 1.5e-07, + "cache_read_input_token_cost_flex": 8e-08, + "cache_read_input_token_cost_priority": 2.7e-07, "input_cost_per_audio_token": 1.5e-06, + "input_cost_per_audio_token_priority": 2.7e-06, "input_cost_per_token": 1.5e-06, + "input_cost_per_token_batches": 7.5e-07, + "input_cost_per_token_flex": 7.5e-07, + "input_cost_per_token_priority": 2.7e-06, "litellm_provider": "gemini", "max_input_tokens": 1048576, "max_output_tokens": 65535, @@ -22552,7 +22575,16 @@ "mode": "chat", "output_cost_per_reasoning_token": 9e-06, "output_cost_per_token": 9e-06, + "output_cost_per_token_batches": 4.5e-06, + "output_cost_per_token_flex": 4.5e-06, + "output_cost_per_token_priority": 1.62e-05, + "prompt_cache_min_tokens": 4096, "rpm": 2000, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/pricing/gemini-3", "supported_endpoints": [ "/v1/chat/completions", @@ -22568,9 +22600,10 @@ "supported_output_modalities": [ "text" ], - "supports_audio_output": false, "supports_audio_input": true, + "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -22582,31 +22615,17 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, "tpm": 800000, - "input_cost_per_token_priority": 2.7e-06, - "input_cost_per_audio_token_priority": 2.7e-06, - "output_cost_per_token_priority": 1.62e-05, - "cache_read_input_token_cost_priority": 2.7e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, - "web_search_billing_unit": "per_query", - "input_cost_per_token_batches": 7.5e-07, - "output_cost_per_token_batches": 4.5e-06, - "input_cost_per_token_flex": 7.5e-07, - "output_cost_per_token_flex": 4.5e-06, - "cache_read_input_token_cost_flex": 8e-08 + "web_search_billing_unit": "per_query" }, "gemini/gemini-3.6-flash": { - "prompt_cache_min_tokens": 4096, "cache_read_input_token_cost": 7.5e-08, "cache_read_input_token_cost_flex": 3.75e-08, + "cache_read_input_token_cost_priority": 1.35e-07, "input_cost_per_token": 7.5e-07, "input_cost_per_token_batches": 3.75e-07, "input_cost_per_token_flex": 3.75e-07, + "input_cost_per_token_priority": 1.35e-06, "litellm_provider": "gemini", "max_input_tokens": 1048576, "max_output_tokens": 65536, @@ -22616,7 +22635,14 @@ "output_cost_per_token": 3.75e-06, "output_cost_per_token_batches": 1.875e-06, "output_cost_per_token_flex": 1.875e-06, + "output_cost_per_token_priority": 6.75e-06, + "prompt_cache_min_tokens": 4096, "rpm": 2000, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/pricing/gemini-3", "supported_endpoints": [ "/v1/chat/completions", @@ -22632,9 +22658,10 @@ "supported_output_modalities": [ "text" ], - "supports_audio_output": false, "supports_audio_input": true, + "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -22646,25 +22673,17 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, "tpm": 800000, - "input_cost_per_token_priority": 1.35e-06, - "output_cost_per_token_priority": 6.75e-06, - "cache_read_input_token_cost_priority": 1.35e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini/gemini-3.7-flash": { - "prompt_cache_min_tokens": 4096, "cache_read_input_token_cost": 7.5e-08, "cache_read_input_token_cost_flex": 3.75e-08, + "cache_read_input_token_cost_priority": 1.35e-07, "input_cost_per_token": 7.5e-07, "input_cost_per_token_batches": 3.75e-07, "input_cost_per_token_flex": 3.75e-07, + "input_cost_per_token_priority": 1.35e-06, "litellm_provider": "gemini", "max_input_tokens": 1048576, "max_output_tokens": 65536, @@ -22674,7 +22693,14 @@ "output_cost_per_token": 3.75e-06, "output_cost_per_token_batches": 1.875e-06, "output_cost_per_token_flex": 1.875e-06, + "output_cost_per_token_priority": 6.75e-06, + "prompt_cache_min_tokens": 4096, "rpm": 2000, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/pricing/gemini-3", "supported_endpoints": [ "/v1/chat/completions", @@ -22690,9 +22716,10 @@ "supported_output_modalities": [ "text" ], - "supports_audio_output": false, "supports_audio_input": true, + "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -22704,16 +22731,7 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, "tpm": 800000, - "input_cost_per_token_priority": 1.35e-06, - "output_cost_per_token_priority": 6.75e-06, - "cache_read_input_token_cost_priority": 1.35e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini/gemini-omni-flash-preview": { @@ -22750,12 +22768,15 @@ "tpm": 800000 }, "gemini/gemini-3.1-pro-preview": { - "prompt_cache_min_tokens": 4096, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, + "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, + "cache_read_input_token_cost_priority": 3.6e-07, "input_cost_per_token": 2e-06, "input_cost_per_token_above_200k_tokens": 4e-06, + "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, "input_cost_per_token_batches": 1e-06, + "input_cost_per_token_priority": 3.6e-06, "litellm_provider": "gemini", "max_input_tokens": 1048576, "max_output_tokens": 65536, @@ -22763,8 +22784,16 @@ "mode": "chat", "output_cost_per_token": 1.2e-05, "output_cost_per_token_above_200k_tokens": 1.8e-05, + "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, "output_cost_per_token_batches": 6e-06, + "output_cost_per_token_priority": 2.16e-05, + "prompt_cache_min_tokens": 4096, "rpm": 2000, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-3.1-pro-preview", "supported_endpoints": [ "/v1/chat/completions", @@ -22782,38 +22811,30 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, + "supports_url_context": true, "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_url_context": true, - "supports_native_streaming": true, "tpm": 800000, - "input_cost_per_token_priority": 3.6e-06, - "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, - "output_cost_per_token_priority": 2.16e-05, - "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, - "cache_read_input_token_cost_priority": 3.6e-07, - "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini/gemini-3.1-pro-preview-customtools": { - "prompt_cache_min_tokens": 4096, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, + "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, + "cache_read_input_token_cost_priority": 3.6e-07, "input_cost_per_token": 2e-06, "input_cost_per_token_above_200k_tokens": 4e-06, + "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, "input_cost_per_token_batches": 1e-06, + "input_cost_per_token_priority": 3.6e-06, "litellm_provider": "gemini", "max_input_tokens": 1048576, "max_output_tokens": 65536, @@ -22821,8 +22842,16 @@ "mode": "chat", "output_cost_per_token": 1.2e-05, "output_cost_per_token_above_200k_tokens": 1.8e-05, + "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, "output_cost_per_token_batches": 6e-06, + "output_cost_per_token_priority": 2.16e-05, + "prompt_cache_min_tokens": 4096, "rpm": 2000, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-3.1-pro-preview", "supported_endpoints": [ "/v1/chat/completions", @@ -22840,35 +22869,27 @@ ], "supports_audio_input": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, + "supports_url_context": true, "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_url_context": true, - "supports_native_streaming": true, "tpm": 800000, - "input_cost_per_token_priority": 3.6e-06, - "input_cost_per_token_above_200k_tokens_priority": 7.2e-06, - "output_cost_per_token_priority": 2.16e-05, - "output_cost_per_token_above_200k_tokens_priority": 3.24e-05, - "cache_read_input_token_cost_priority": 3.6e-07, - "cache_read_input_token_cost_above_200k_tokens_priority": 7.2e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini-3-flash-preview": { "cache_read_input_token_cost": 5e-08, + "cache_read_input_token_cost_priority": 9e-08, "input_cost_per_audio_token": 1e-06, + "input_cost_per_audio_token_priority": 1.8e-06, "input_cost_per_token": 5e-07, + "input_cost_per_token_priority": 9e-07, "litellm_provider": "vertex_ai-language-models", "max_input_tokens": 1048576, "max_output_tokens": 65535, @@ -22876,6 +22897,12 @@ "mode": "chat", "output_cost_per_reasoning_token": 3e-06, "output_cost_per_token": 3e-06, + "output_cost_per_token_priority": 5.4e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/pricing/gemini-3", "supported_endpoints": [ "/v1/chat/completions", @@ -22893,6 +22920,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -22903,16 +22931,6 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 9e-07, - "input_cost_per_audio_token_priority": 1.8e-06, - "output_cost_per_token_priority": 5.4e-06, - "cache_read_input_token_cost_priority": 9e-08, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini-omni-flash-preview": { @@ -22947,11 +22965,16 @@ "supports_vision": true }, "gemini-3.5-flash": { - "prompt_cache_min_tokens": 4096, - "deprecation_date": "2027-05-19", "cache_read_input_token_cost": 1.5e-07, + "cache_read_input_token_cost_flex": 7.5e-08, + "cache_read_input_token_cost_priority": 2.7e-07, + "deprecation_date": "2027-05-19", "input_cost_per_audio_token": 1.5e-06, + "input_cost_per_audio_token_priority": 2.7e-06, "input_cost_per_token": 1.5e-06, + "input_cost_per_token_batches": 7.5e-07, + "input_cost_per_token_flex": 7.5e-07, + "input_cost_per_token_priority": 2.7e-06, "litellm_provider": "vertex_ai-language-models", "max_input_tokens": 1048576, "max_output_tokens": 65535, @@ -22959,6 +22982,15 @@ "mode": "chat", "output_cost_per_reasoning_token": 9e-06, "output_cost_per_token": 9e-06, + "output_cost_per_token_batches": 4.5e-06, + "output_cost_per_token_flex": 4.5e-06, + "output_cost_per_token_priority": 1.62e-05, + "prompt_cache_min_tokens": 4096, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/pricing/gemini-3", "supported_endpoints": [ "/v1/chat/completions", @@ -22974,9 +23006,10 @@ "supported_output_modalities": [ "text" ], - "supports_audio_output": false, "supports_audio_input": true, + "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -22988,30 +23021,16 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 2.7e-06, - "input_cost_per_audio_token_priority": 2.7e-06, - "output_cost_per_token_priority": 1.62e-05, - "cache_read_input_token_cost_priority": 2.7e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, - "web_search_billing_unit": "per_query", - "input_cost_per_token_batches": 7.5e-07, - "output_cost_per_token_batches": 4.5e-06, - "input_cost_per_token_flex": 7.5e-07, - "output_cost_per_token_flex": 4.5e-06, - "cache_read_input_token_cost_flex": 7.5e-08 + "web_search_billing_unit": "per_query" }, "gemini-3.6-flash": { - "prompt_cache_min_tokens": 4096, "cache_read_input_token_cost": 7.5e-08, "cache_read_input_token_cost_flex": 3.75e-08, + "cache_read_input_token_cost_priority": 1.35e-07, "input_cost_per_token": 7.5e-07, "input_cost_per_token_batches": 3.75e-07, "input_cost_per_token_flex": 3.75e-07, + "input_cost_per_token_priority": 1.35e-06, "litellm_provider": "vertex_ai-language-models", "max_input_tokens": 1048576, "max_output_tokens": 65536, @@ -23021,6 +23040,13 @@ "output_cost_per_token": 3.75e-06, "output_cost_per_token_batches": 1.875e-06, "output_cost_per_token_flex": 1.875e-06, + "output_cost_per_token_priority": 6.75e-06, + "prompt_cache_min_tokens": 4096, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/pricing/gemini-3", "supported_endpoints": [ "/v1/chat/completions", @@ -23036,9 +23062,10 @@ "supported_output_modalities": [ "text" ], - "supports_audio_output": false, "supports_audio_input": true, + "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -23050,24 +23077,16 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 1.35e-06, - "output_cost_per_token_priority": 6.75e-06, - "cache_read_input_token_cost_priority": 1.35e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini-3.7-flash": { - "prompt_cache_min_tokens": 4096, "cache_read_input_token_cost": 7.5e-08, "cache_read_input_token_cost_flex": 3.75e-08, + "cache_read_input_token_cost_priority": 1.35e-07, "input_cost_per_token": 7.5e-07, "input_cost_per_token_batches": 3.75e-07, "input_cost_per_token_flex": 3.75e-07, + "input_cost_per_token_priority": 1.35e-06, "litellm_provider": "vertex_ai-language-models", "max_input_tokens": 1048576, "max_output_tokens": 65536, @@ -23077,6 +23096,13 @@ "output_cost_per_token": 3.75e-06, "output_cost_per_token_batches": 1.875e-06, "output_cost_per_token_flex": 1.875e-06, + "output_cost_per_token_priority": 6.75e-06, + "prompt_cache_min_tokens": 4096, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://ai.google.dev/pricing/gemini-3", "supported_endpoints": [ "/v1/chat/completions", @@ -23092,9 +23118,10 @@ "supported_output_modalities": [ "text" ], - "supports_audio_output": false, "supports_audio_input": true, + "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -23106,15 +23133,6 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "input_cost_per_token_priority": 1.35e-06, - "output_cost_per_token_priority": 6.75e-06, - "cache_read_input_token_cost_priority": 1.35e-07, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "gemini/gemini-2.5-pro-preview-tts": { @@ -23131,6 +23149,11 @@ "output_cost_per_token": 1e-05, "output_cost_per_token_above_200k_tokens": 1.5e-05, "rpm": 10000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-2.5-pro-preview", "supported_modalities": [ "text" @@ -23146,12 +23169,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "tpm": 10000000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 10000000 }, "gemini/gemini-exp-1114": { "input_cost_per_token": 0, @@ -23206,12 +23224,12 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.05e-06, + "rpm": 10, "source": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/models#foundation_models", "supports_function_calling": true, "supports_tool_choice": true, "supports_vision": true, - "tpm": 250000, - "rpm": 10 + "tpm": 250000 }, "gemini/gemini-gemma-2-9b-it": { "input_cost_per_token": 3.5e-07, @@ -23220,12 +23238,12 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.05e-06, + "rpm": 10, "source": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/models#foundation_models", "supports_function_calling": true, "supports_tool_choice": true, "supports_vision": true, - "tpm": 250000, - "rpm": 10 + "tpm": 250000 }, "gemini/gemma-3-27b-it": { "input_cost_per_audio_per_second": 0, @@ -23459,22 +23477,22 @@ ] }, "github_copilot/claude-haiku-4.5": { + "cache_creation_input_token_cost": 1.25e-06, + "cache_read_input_token_cost": 1e-07, + "input_cost_per_token": 1e-06, "litellm_provider": "github_copilot", "max_input_tokens": 128000, "max_output_tokens": 16000, "max_tokens": 16000, "mode": "chat", + "output_cost_per_token": 5e-06, + "source": "https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing", "supported_endpoints": [ "/v1/chat/completions" ], "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, - "cache_creation_input_token_cost": 1.25e-06, - "cache_read_input_token_cost": 1e-07, - "input_cost_per_token": 1e-06, - "output_cost_per_token": 5e-06, - "source": "https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing" + "supports_vision": true }, "github_copilot/claude-opus-4.5": { "litellm_provider": "github_copilot", @@ -23486,9 +23504,24 @@ "/v1/chat/completions" ], "supports_function_calling": true, + "supports_output_config": true, "supports_parallel_function_calling": true, - "supports_vision": true, - "supports_output_config": true + "supports_vision": true + }, + "github_copilot/claude-opus-4.6-fast": { + "supports_adaptive_thinking": true, + "supports_legacy_thinking": true, + "litellm_provider": "github_copilot", + "max_input_tokens": 128000, + "max_output_tokens": 16000, + "max_tokens": 16000, + "mode": "chat", + "supported_endpoints": [ + "/v1/chat/completions" + ], + "supports_function_calling": true, + "supports_parallel_function_calling": true, + "supports_vision": true }, "github_copilot/claude-sonnet-4.5": { "litellm_provider": "github_copilot", @@ -23616,38 +23649,38 @@ "supports_parallel_function_calling": true }, "github_copilot/gpt-5-mini": { + "cache_read_input_token_cost": 2.5e-08, + "input_cost_per_token": 2.5e-07, "litellm_provider": "github_copilot", "max_input_tokens": 128000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", + "output_cost_per_token": 2e-06, + "source": "https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing", "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_response_schema": true, - "supports_vision": true, - "cache_read_input_token_cost": 2.5e-08, - "input_cost_per_token": 2.5e-07, - "output_cost_per_token": 2e-06, - "source": "https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing" + "supports_vision": true }, "github_copilot/gpt-5.3-codex": { + "cache_read_input_token_cost": 1.75e-07, + "input_cost_per_token": 1.75e-06, "litellm_provider": "github_copilot", "max_input_tokens": 128000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "responses", + "output_cost_per_token": 1.4e-05, + "source": "https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing", "supported_endpoints": [ "/v1/responses" ], "supports_function_calling": true, "supports_parallel_function_calling": true, + "supports_reasoning": true, "supports_response_schema": true, - "supports_vision": true, - "cache_read_input_token_cost": 1.75e-07, - "input_cost_per_token": 1.75e-06, - "output_cost_per_token": 1.4e-05, - "source": "https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing", - "supports_reasoning": true + "supports_vision": true }, "github_copilot/mai-code-1-flash": { "cache_read_input_token_cost": 7.5e-08, @@ -23914,8 +23947,8 @@ "mode": "chat", "output_cost_per_token": 2.5e-05, "supports_function_calling": true, - "supports_vision": true, - "supports_output_config": true + "supports_output_config": true, + "supports_vision": true }, "gmi/anthropic/claude-sonnet-4.5": { "input_cost_per_token": 3e-06, @@ -24188,20 +24221,22 @@ "global.anthropic.claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, + "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.2e-05, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.2e-05, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "input_cost_per_token_batches": 1.5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "output_cost_per_token_batches": 7.5e-06, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -24210,33 +24245,33 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024, - "input_cost_per_token_batches": 1.5e-06, - "output_cost_per_token_batches": 7.5e-06 + "supports_tool_search": true, + "supports_vision": true }, "global.anthropic.claude-sonnet-4-20250514-v1:0": { + "bedrock_converse_supports_strict_tools": false, "cache_creation_input_token_cost": 3.75e-06, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, "deprecation_date": "2026-10-14", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -24250,37 +24285,35 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "bedrock_converse_supports_strict_tools": false, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "global.anthropic.claude-haiku-4-5-20251001-v1:0": { "cache_creation_input_token_cost": 1.25e-06, "cache_creation_input_token_cost_above_1hr": 2e-06, "cache_read_input_token_cost": 1e-07, "input_cost_per_token": 1e-06, + "input_cost_per_token_batches": 5e-07, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, + "output_cost_per_token_batches": 2.5e-06, + "prompt_cache_min_tokens": 4096, "source": "https://aws.amazon.com/about-aws/whats-new/2025/10/claude-4-5-haiku-anthropic-amazon-bedrock", "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096, - "input_cost_per_token_batches": 5e-07, - "output_cost_per_token_batches": 2.5e-06 + "supports_tool_search": true, + "supports_vision": true }, "global.amazon.nova-2-lite-v1:0": { "cache_read_input_token_cost": 7.5e-08, @@ -24541,16 +24574,16 @@ "cache_read_input_token_cost": 5e-07, "cache_read_input_token_cost_priority": 8.75e-07, "input_cost_per_token": 2e-06, - "input_cost_per_token_priority": 3.5e-06, "input_cost_per_token_batches": 1e-06, + "input_cost_per_token_priority": 3.5e-06, "litellm_provider": "openai", "max_input_tokens": 1047576, "max_output_tokens": 32768, "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 8e-06, - "output_cost_per_token_priority": 1.4e-05, "output_cost_per_token_batches": 4e-06, + "output_cost_per_token_priority": 1.4e-05, "search_context_cost_per_query": { "search_context_size_high": 0.025, "search_context_size_low": 0.025, @@ -24625,16 +24658,16 @@ "cache_read_input_token_cost": 1e-07, "cache_read_input_token_cost_priority": 1.75e-07, "input_cost_per_token": 4e-07, - "input_cost_per_token_priority": 7e-07, "input_cost_per_token_batches": 2e-07, + "input_cost_per_token_priority": 7e-07, "litellm_provider": "openai", "max_input_tokens": 1047576, "max_output_tokens": 32768, "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 1.6e-06, - "output_cost_per_token_priority": 2.8e-06, "output_cost_per_token_batches": 8e-07, + "output_cost_per_token_priority": 2.8e-06, "search_context_cost_per_query": { "search_context_size_high": 0.025, "search_context_size_low": 0.025, @@ -24705,16 +24738,16 @@ "cache_read_input_token_cost_priority": 5e-08, "deprecation_date": "2026-10-23", "input_cost_per_token": 1e-07, - "input_cost_per_token_priority": 2e-07, "input_cost_per_token_batches": 5e-08, + "input_cost_per_token_priority": 2e-07, "litellm_provider": "openai", "max_input_tokens": 1047576, "max_output_tokens": 32768, "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 4e-07, - "output_cost_per_token_priority": 8e-07, "output_cost_per_token_batches": 2e-07, + "output_cost_per_token_priority": 8e-07, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -24785,16 +24818,16 @@ "cache_read_input_token_cost": 1.25e-06, "cache_read_input_token_cost_priority": 2.125e-06, "input_cost_per_token": 2.5e-06, - "input_cost_per_token_priority": 4.25e-06, "input_cost_per_token_batches": 1.25e-06, + "input_cost_per_token_priority": 4.25e-06, "litellm_provider": "openai", "max_input_tokens": 128000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", "output_cost_per_token": 1e-05, - "output_cost_per_token_priority": 1.7e-05, "output_cost_per_token_batches": 5e-06, + "output_cost_per_token_priority": 1.7e-05, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, @@ -24808,16 +24841,16 @@ "cache_read_input_token_cost": 1.25e-06, "cache_read_input_token_cost_priority": 2.125e-06, "input_cost_per_token": 2.5e-06, - "input_cost_per_token_priority": 4.25e-06, "input_cost_per_token_batches": 1.25e-06, + "input_cost_per_token_priority": 4.25e-06, "litellm_provider": "openai", "max_input_tokens": 128000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", "output_cost_per_token": 1e-05, - "output_cost_per_token_priority": 1.7e-05, "output_cost_per_token_batches": 5e-06, + "output_cost_per_token_priority": 1.7e-05, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, @@ -25063,6 +25096,7 @@ "supports_vision": false }, "gpt-audio-mini-2025-12-15": { + "deprecation_date": "2027-01-20", "input_cost_per_audio_token": 1e-05, "input_cost_per_token": 6e-07, "litellm_provider": "openai", @@ -25096,8 +25130,7 @@ "supports_response_schema": false, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": false, - "deprecation_date": "2027-01-20" + "supports_vision": false }, "gpt-4o-mini": { "cache_read_input_token_cost": 7.5e-08, @@ -25126,16 +25159,16 @@ "cache_read_input_token_cost": 7.5e-08, "cache_read_input_token_cost_priority": 1.25e-07, "input_cost_per_token": 1.5e-07, - "input_cost_per_token_priority": 2.5e-07, "input_cost_per_token_batches": 7.5e-08, + "input_cost_per_token_priority": 2.5e-07, "litellm_provider": "openai", "max_input_tokens": 128000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", "output_cost_per_token": 6e-07, - "output_cost_per_token_priority": 1e-06, "output_cost_per_token_batches": 3e-07, + "output_cost_per_token_priority": 1e-06, "search_context_cost_per_query": { "search_context_size_high": 0.03, "search_context_size_low": 0.025, @@ -25432,62 +25465,62 @@ "gpt-image-1.5": { "cache_read_input_token_cost": 1.25e-06, "deprecation_date": "2026-12-01", + "input_cost_per_image_token": 8e-06, "input_cost_per_token": 5e-06, "litellm_provider": "openai", "mode": "image_generation", - "output_cost_per_token": 1e-05, - "input_cost_per_image_token": 8e-06, "output_cost_per_image_token": 3.2e-05, + "output_cost_per_token": 1e-05, "supported_endpoints": [ "/v1/images/generations" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "gpt-image-1.5-2025-12-16": { "cache_read_input_token_cost": 1.25e-06, "deprecation_date": "2026-12-01", + "input_cost_per_image_token": 8e-06, "input_cost_per_token": 5e-06, "litellm_provider": "openai", "mode": "image_generation", - "output_cost_per_token": 1e-05, - "input_cost_per_image_token": 8e-06, "output_cost_per_image_token": 3.2e-05, + "output_cost_per_token": 1e-05, "supported_endpoints": [ "/v1/images/generations" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "gpt-image-2": { "cache_read_input_token_cost": 1.25e-06, + "input_cost_per_image_token": 8e-06, "input_cost_per_token": 5e-06, "litellm_provider": "openai", "mode": "image_generation", - "output_cost_per_token": 1e-05, - "input_cost_per_image_token": 8e-06, "output_cost_per_image_token": 3e-05, + "output_cost_per_token": 1e-05, "supported_endpoints": [ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "gpt-image-2-2026-04-21": { "cache_read_input_token_cost": 1.25e-06, + "input_cost_per_image_token": 8e-06, "input_cost_per_token": 5e-06, "litellm_provider": "openai", "mode": "image_generation", - "output_cost_per_token": 1e-05, - "input_cost_per_image_token": 8e-06, "output_cost_per_image_token": 3e-05, + "output_cost_per_token": 1e-05, "supported_endpoints": [ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "low/1024-x-1024/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25498,8 +25531,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "low/1024-x-1536/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25510,8 +25543,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "low/1536-x-1024/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25522,8 +25555,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "medium/1024-x-1024/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25534,8 +25567,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "medium/1024-x-1536/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25546,8 +25579,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "medium/1536-x-1024/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25558,8 +25591,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "high/1024-x-1024/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25570,8 +25603,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "high/1024-x-1536/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25582,8 +25615,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "high/1536-x-1024/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25594,8 +25627,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "standard/1024-x-1024/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25606,8 +25639,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "standard/1024-x-1536/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25618,8 +25651,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "standard/1536-x-1024/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25630,8 +25663,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "1024-x-1024/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25642,8 +25675,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "1024-x-1536/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25654,8 +25687,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "1536-x-1024/gpt-image-1.5": { "deprecation_date": "2026-12-01", @@ -25666,8 +25699,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "low/1024-x-1024/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.009, @@ -25677,8 +25710,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "low/1024-x-1536/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.013, @@ -25688,8 +25721,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "low/1536-x-1024/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.013, @@ -25699,8 +25732,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "medium/1024-x-1024/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.034, @@ -25710,8 +25743,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "medium/1024-x-1536/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.05, @@ -25721,8 +25754,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "medium/1536-x-1024/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.05, @@ -25732,8 +25765,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "high/1024-x-1024/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.133, @@ -25743,8 +25776,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "high/1024-x-1536/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.2, @@ -25754,8 +25787,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "high/1536-x-1024/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.2, @@ -25765,8 +25798,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "standard/1024-x-1024/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.009, @@ -25776,8 +25809,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "standard/1024-x-1536/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.013, @@ -25787,8 +25820,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "standard/1536-x-1024/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.013, @@ -25798,8 +25831,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "1024-x-1024/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.009, @@ -25809,8 +25842,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "1024-x-1536/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.013, @@ -25820,8 +25853,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "1536-x-1024/gpt-image-1.5-2025-12-16": { "input_cost_per_image": 0.013, @@ -25831,8 +25864,8 @@ "/v1/images/generations", "/v1/images/edits" ], - "supports_vision": true, - "supports_pdf_input": true + "supports_pdf_input": true, + "supports_vision": true }, "gpt-5": { "cache_read_input_token_cost": 1.25e-07, @@ -25867,7 +25900,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -25877,9 +25912,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5.1": { "cache_read_input_token_cost": 1.25e-07, @@ -25911,7 +25944,9 @@ "image" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -25921,9 +25956,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5.1-2025-11-13": { "cache_read_input_token_cost": 1.25e-07, @@ -25955,7 +25988,9 @@ "image" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -25965,9 +26000,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5.1-chat-latest": { "cache_read_input_token_cost": 1.25e-07, @@ -26000,7 +26033,9 @@ "image" ], "supports_function_calling": false, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": false, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26010,9 +26045,7 @@ "supports_tool_choice": false, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5.2": { "cache_read_input_token_cost": 1.75e-07, @@ -26045,7 +26078,9 @@ "image" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26055,9 +26090,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "gpt-5.2-2025-12-11": { "cache_read_input_token_cost": 1.75e-07, @@ -26090,7 +26123,9 @@ "image" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26100,9 +26135,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "gpt-5.2-chat-latest": { "cache_read_input_token_cost": 1.75e-07, @@ -26134,7 +26167,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26144,9 +26179,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5.3-chat-latest": { "cache_read_input_token_cost": 1.75e-07, @@ -26178,7 +26211,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26188,9 +26223,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5.2-pro": { "input_cost_per_token": 2.1e-05, @@ -26217,7 +26250,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26227,9 +26262,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "gpt-5.2-pro-2025-12-11": { "input_cost_per_token": 2.1e-05, @@ -26256,7 +26289,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26266,9 +26301,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "gpt-5.6": { "cache_creation_input_token_cost": 5e-06, @@ -26368,6 +26401,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://developers.openai.com/api/docs/models/gpt-5.6-sol", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -26391,13 +26425,12 @@ "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_xhigh_reasoning_effort": true, - "supports_service_tier": true, - "source": "https://developers.openai.com/api/docs/models/gpt-5.6-sol" + "supports_xhigh_reasoning_effort": true }, "gpt-5.6-terra": { "cache_creation_input_token_cost": 2.5e-06, @@ -26434,6 +26467,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://developers.openai.com/api/docs/models/gpt-5.6-terra", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -26456,13 +26490,12 @@ "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_xhigh_reasoning_effort": true, - "supports_service_tier": true, - "source": "https://developers.openai.com/api/docs/models/gpt-5.6-terra" + "supports_xhigh_reasoning_effort": true }, "gpt-5.6-luna": { "cache_creation_input_token_cost": 2.5e-07, @@ -26499,6 +26532,7 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "source": "https://developers.openai.com/api/docs/models/gpt-5.6-luna", "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -26521,13 +26555,12 @@ "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_xhigh_reasoning_effort": true, - "supports_service_tier": true, - "source": "https://developers.openai.com/api/docs/models/gpt-5.6-luna" + "supports_xhigh_reasoning_effort": true }, "gpt-5.6-cyber": { "cache_creation_input_token_cost": 1.5625e-05, @@ -26543,6 +26576,7 @@ "mode": "chat", "output_cost_per_token": 7.5e-05, "output_cost_per_token_above_272k_tokens": 0.0001125, + "source": "https://platform.openai.com/docs/models/gpt-5.6-cyber", "supported_endpoints": [ "/v1/responses" ], @@ -26553,8 +26587,10 @@ "supported_output_modalities": [ "text" ], + "supports_computer_use": true, "supports_function_calling": true, "supports_native_streaming": true, + "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -26562,10 +26598,7 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "source": "https://platform.openai.com/docs/models/gpt-5.6-cyber", - "supports_computer_use": true, - "supports_parallel_function_calling": true + "supports_web_search": true }, "daybreak-red-latest": { "cache_creation_input_token_cost": 1.5625e-05, @@ -26581,6 +26614,7 @@ "mode": "chat", "output_cost_per_token": 7.5e-05, "output_cost_per_token_above_272k_tokens": 0.0001125, + "source": "https://platform.openai.com/docs/models/daybreak-red-latest", "supported_endpoints": [ "/v1/responses" ], @@ -26591,8 +26625,10 @@ "supported_output_modalities": [ "text" ], + "supports_computer_use": true, "supports_function_calling": true, "supports_native_streaming": true, + "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -26600,10 +26636,7 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "source": "https://platform.openai.com/docs/models/daybreak-red-latest", - "supports_computer_use": true, - "supports_parallel_function_calling": true + "supports_web_search": true }, "daybreak-blue-latest": { "cache_creation_input_token_cost": 5e-06, @@ -26619,6 +26652,7 @@ "mode": "chat", "output_cost_per_token": 2e-05, "output_cost_per_token_above_272k_tokens": 3e-05, + "source": "https://platform.openai.com/docs/models/daybreak-blue-latest", "supported_endpoints": [ "/v1/responses" ], @@ -26632,6 +26666,7 @@ "supports_computer_use": true, "supports_function_calling": true, "supports_native_streaming": true, + "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -26639,9 +26674,7 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "source": "https://platform.openai.com/docs/models/daybreak-blue-latest", - "supports_parallel_function_calling": true + "supports_web_search": true }, "chat-latest": { "cache_read_input_token_cost": 5e-07, @@ -26652,6 +26685,11 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 3e-05, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, "source": "https://platform.openai.com/docs/models/chat-latest", "supported_endpoints": [ "/v1/chat/completions", @@ -26664,6 +26702,7 @@ "supported_output_modalities": [ "text" ], + "supports_file_search": true, "supports_function_calling": true, "supports_native_streaming": true, "supports_parallel_function_calling": true, @@ -26673,13 +26712,7 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "search_context_cost_per_query": { - "search_context_size_high": 0.01, - "search_context_size_low": 0.01, - "search_context_size_medium": 0.01 - }, - "supports_file_search": true + "supports_web_search": true }, "gpt-5.5": { "cache_read_input_token_cost": 5e-07, @@ -26688,8 +26721,8 @@ "cache_read_input_token_cost_priority": 1e-06, "input_cost_per_token": 5e-06, "input_cost_per_token_above_272k_tokens": 1e-05, - "input_cost_per_token_flex": 2.5e-06, "input_cost_per_token_batches": 2.5e-06, + "input_cost_per_token_flex": 2.5e-06, "input_cost_per_token_priority": 1e-05, "litellm_provider": "openai", "max_input_tokens": 1050000, @@ -26698,8 +26731,8 @@ "mode": "chat", "output_cost_per_token": 3e-05, "output_cost_per_token_above_272k_tokens": 4.5e-05, - "output_cost_per_token_flex": 1.5e-05, "output_cost_per_token_batches": 1.5e-05, + "output_cost_per_token_flex": 1.5e-05, "output_cost_per_token_priority": 6e-05, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, @@ -26721,7 +26754,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26731,9 +26766,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "gpt-5.5-2026-04-23": { "cache_read_input_token_cost": 5e-07, @@ -26742,8 +26775,8 @@ "cache_read_input_token_cost_priority": 1e-06, "input_cost_per_token": 5e-06, "input_cost_per_token_above_272k_tokens": 1e-05, - "input_cost_per_token_flex": 2.5e-06, "input_cost_per_token_batches": 2.5e-06, + "input_cost_per_token_flex": 2.5e-06, "input_cost_per_token_priority": 1e-05, "litellm_provider": "openai", "max_input_tokens": 1050000, @@ -26752,8 +26785,8 @@ "mode": "chat", "output_cost_per_token": 3e-05, "output_cost_per_token_above_272k_tokens": 4.5e-05, - "output_cost_per_token_flex": 1.5e-05, "output_cost_per_token_batches": 1.5e-05, + "output_cost_per_token_flex": 1.5e-05, "output_cost_per_token_priority": 6e-05, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, @@ -26775,7 +26808,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26785,17 +26820,15 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "gpt-5.5-pro": { "cache_read_input_token_cost": 3e-06, "cache_read_input_token_cost_above_272k_tokens": 6e-06, "input_cost_per_token": 3e-05, "input_cost_per_token_above_272k_tokens": 6e-05, - "input_cost_per_token_flex": 1.5e-05, "input_cost_per_token_batches": 1.5e-05, + "input_cost_per_token_flex": 1.5e-05, "litellm_provider": "openai", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -26803,8 +26836,8 @@ "mode": "responses", "output_cost_per_token": 0.00018, "output_cost_per_token_above_272k_tokens": 0.00027, - "output_cost_per_token_flex": 9e-05, "output_cost_per_token_batches": 9e-05, + "output_cost_per_token_flex": 9e-05, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, "search_context_cost_per_query": { @@ -26824,7 +26857,10 @@ "text" ], "supports_function_calling": true, + "supports_low_reasoning_effort": false, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26834,18 +26870,15 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "supports_low_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "gpt-5.5-pro-2026-04-23": { "cache_read_input_token_cost": 3e-06, "cache_read_input_token_cost_above_272k_tokens": 6e-06, "input_cost_per_token": 3e-05, "input_cost_per_token_above_272k_tokens": 6e-05, - "input_cost_per_token_flex": 1.5e-05, "input_cost_per_token_batches": 1.5e-05, + "input_cost_per_token_flex": 1.5e-05, "litellm_provider": "openai", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -26853,8 +26886,8 @@ "mode": "responses", "output_cost_per_token": 0.00018, "output_cost_per_token_above_272k_tokens": 0.00027, - "output_cost_per_token_flex": 9e-05, "output_cost_per_token_batches": 9e-05, + "output_cost_per_token_flex": 9e-05, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, "search_context_cost_per_query": { @@ -26874,7 +26907,10 @@ "text" ], "supports_function_calling": true, + "supports_low_reasoning_effort": false, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26884,10 +26920,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "supports_low_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "gpt-5.4": { "cache_read_input_token_cost": 2.5e-07, @@ -26896,8 +26929,8 @@ "cache_read_input_token_cost_priority": 5e-07, "input_cost_per_token": 2.5e-06, "input_cost_per_token_above_272k_tokens": 5e-06, - "input_cost_per_token_flex": 1.25e-06, "input_cost_per_token_batches": 1.25e-06, + "input_cost_per_token_flex": 1.25e-06, "input_cost_per_token_priority": 5e-06, "litellm_provider": "openai", "max_input_tokens": 1050000, @@ -26906,8 +26939,8 @@ "mode": "chat", "output_cost_per_token": 1.5e-05, "output_cost_per_token_above_272k_tokens": 2.25e-05, - "output_cost_per_token_flex": 7.5e-06, "output_cost_per_token_batches": 7.5e-06, + "output_cost_per_token_flex": 7.5e-06, "output_cost_per_token_priority": 3e-05, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, @@ -26924,7 +26957,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26933,9 +26968,7 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "gpt-5.4-2026-03-05": { "cache_read_input_token_cost": 2.5e-07, @@ -26944,8 +26977,8 @@ "cache_read_input_token_cost_priority": 5e-07, "input_cost_per_token": 2.5e-06, "input_cost_per_token_above_272k_tokens": 5e-06, - "input_cost_per_token_flex": 1.25e-06, "input_cost_per_token_batches": 1.25e-06, + "input_cost_per_token_flex": 1.25e-06, "input_cost_per_token_priority": 5e-06, "litellm_provider": "openai", "max_input_tokens": 1050000, @@ -26954,8 +26987,8 @@ "mode": "chat", "output_cost_per_token": 1.5e-05, "output_cost_per_token_above_272k_tokens": 2.25e-05, - "output_cost_per_token_flex": 7.5e-06, "output_cost_per_token_batches": 7.5e-06, + "output_cost_per_token_flex": 7.5e-06, "output_cost_per_token_priority": 3e-05, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, @@ -26972,7 +27005,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -26981,17 +27016,15 @@ "supports_system_messages": true, "supports_tool_choice": true, "supports_vision": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "gpt-5.4-pro": { "cache_read_input_token_cost": 3e-06, "cache_read_input_token_cost_above_272k_tokens": 6e-06, "input_cost_per_token": 3e-05, "input_cost_per_token_above_272k_tokens": 6e-05, - "input_cost_per_token_flex": 1.5e-05, "input_cost_per_token_batches": 1.5e-05, + "input_cost_per_token_flex": 1.5e-05, "litellm_provider": "openai", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -26999,8 +27032,8 @@ "mode": "responses", "output_cost_per_token": 0.00018, "output_cost_per_token_above_272k_tokens": 0.00027, - "output_cost_per_token_flex": 9e-05, "output_cost_per_token_batches": 9e-05, + "output_cost_per_token_flex": 9e-05, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, "search_context_cost_per_query": { @@ -27020,7 +27053,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27030,17 +27065,15 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "gpt-5.4-pro-2026-03-05": { "cache_read_input_token_cost": 3e-06, "cache_read_input_token_cost_above_272k_tokens": 6e-06, "input_cost_per_token": 3e-05, "input_cost_per_token_above_272k_tokens": 6e-05, - "input_cost_per_token_flex": 1.5e-05, "input_cost_per_token_batches": 1.5e-05, + "input_cost_per_token_flex": 1.5e-05, "litellm_provider": "openai", "max_input_tokens": 1050000, "max_output_tokens": 128000, @@ -27048,8 +27081,8 @@ "mode": "responses", "output_cost_per_token": 0.00018, "output_cost_per_token_above_272k_tokens": 0.00027, - "output_cost_per_token_flex": 9e-05, "output_cost_per_token_batches": 9e-05, + "output_cost_per_token_flex": 9e-05, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, "search_context_cost_per_query": { @@ -27069,7 +27102,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27079,17 +27114,15 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "gpt-5.4-mini": { "cache_read_input_token_cost": 7.5e-08, "cache_read_input_token_cost_flex": 3.75e-08, "cache_read_input_token_cost_priority": 1.5e-07, "input_cost_per_token": 7.5e-07, - "input_cost_per_token_flex": 3.75e-07, "input_cost_per_token_batches": 3.75e-07, + "input_cost_per_token_flex": 3.75e-07, "input_cost_per_token_priority": 1.5e-06, "litellm_provider": "openai", "max_input_tokens": 272000, @@ -27097,8 +27130,8 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 4.5e-06, - "output_cost_per_token_flex": 2.25e-06, "output_cost_per_token_batches": 2.25e-06, + "output_cost_per_token_flex": 2.25e-06, "output_cost_per_token_priority": 9e-06, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, @@ -27120,7 +27153,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27130,17 +27165,15 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "gpt-5.4-mini-2026-03-17": { "cache_read_input_token_cost": 7.5e-08, "cache_read_input_token_cost_flex": 3.75e-08, "cache_read_input_token_cost_priority": 1.5e-07, "input_cost_per_token": 7.5e-07, - "input_cost_per_token_flex": 3.75e-07, "input_cost_per_token_batches": 3.75e-07, + "input_cost_per_token_flex": 3.75e-07, "input_cost_per_token_priority": 1.5e-06, "litellm_provider": "openai", "max_input_tokens": 272000, @@ -27148,8 +27181,8 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 4.5e-06, - "output_cost_per_token_flex": 2.25e-06, "output_cost_per_token_batches": 2.25e-06, + "output_cost_per_token_flex": 2.25e-06, "output_cost_per_token_priority": 9e-06, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, @@ -27171,7 +27204,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27181,24 +27216,22 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "gpt-5.4-nano": { "cache_read_input_token_cost": 2e-08, "cache_read_input_token_cost_flex": 1e-08, "input_cost_per_token": 2e-07, - "input_cost_per_token_flex": 1e-07, "input_cost_per_token_batches": 1e-07, + "input_cost_per_token_flex": 1e-07, "litellm_provider": "openai", "max_input_tokens": 272000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1.25e-06, - "output_cost_per_token_flex": 6.25e-07, "output_cost_per_token_batches": 6.25e-07, + "output_cost_per_token_flex": 6.25e-07, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, "search_context_cost_per_query": { @@ -27219,7 +27252,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27229,24 +27264,22 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "gpt-5.4-nano-2026-03-17": { "cache_read_input_token_cost": 2e-08, "cache_read_input_token_cost_flex": 1e-08, "input_cost_per_token": 2e-07, - "input_cost_per_token_flex": 1e-07, "input_cost_per_token_batches": 1e-07, + "input_cost_per_token_flex": 1e-07, "litellm_provider": "openai", "max_input_tokens": 272000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1.25e-06, - "output_cost_per_token_flex": 6.25e-07, "output_cost_per_token_batches": 6.25e-07, + "output_cost_per_token_flex": 6.25e-07, "regional_processing_uplift_multiplier_eu": 1.1, "regional_processing_uplift_multiplier_us": 1.1, "search_context_cost_per_query": { @@ -27267,7 +27300,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27277,9 +27312,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": true, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false + "supports_xhigh_reasoning_effort": true }, "gpt-5-pro": { "input_cost_per_token": 1.5e-05, @@ -27308,7 +27341,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": false, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27318,9 +27353,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5-pro-2025-10-06": { "deprecation_date": "2026-12-11", @@ -27350,7 +27383,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": false, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27360,9 +27395,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5-2025-08-07": { "cache_read_input_token_cost": 1.25e-07, @@ -27398,7 +27431,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27408,9 +27443,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5-chat": { "cache_read_input_token_cost": 1.25e-07, @@ -27434,7 +27467,9 @@ "text" ], "supports_function_calling": false, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": false, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27443,9 +27478,7 @@ "supports_system_messages": true, "supports_tool_choice": false, "supports_vision": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5-chat-latest": { "cache_read_input_token_cost": 1.25e-07, @@ -27470,7 +27503,9 @@ "text" ], "supports_function_calling": false, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": false, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27479,9 +27514,7 @@ "supports_system_messages": true, "supports_tool_choice": false, "supports_vision": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5-codex": { "cache_read_input_token_cost": 1.25e-07, @@ -27509,7 +27542,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27519,9 +27554,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5.1-codex": { "cache_read_input_token_cost": 1.25e-07, @@ -27552,7 +27585,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27562,9 +27597,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5.1-codex-max": { "cache_read_input_token_cost": 1.25e-07, @@ -27592,7 +27625,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27602,9 +27637,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "gpt-5.1-codex-mini": { "cache_read_input_token_cost": 2.5e-08, @@ -27635,7 +27668,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27645,9 +27680,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5.2-codex": { "cache_read_input_token_cost": 1.75e-07, @@ -27678,7 +27711,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27688,9 +27723,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": true }, "gpt-5.3-codex": { "cache_read_input_token_cost": 1.75e-07, @@ -27720,7 +27753,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27730,9 +27765,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5-mini": { "cache_read_input_token_cost": 2.5e-08, @@ -27767,7 +27800,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27777,9 +27812,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5-mini-2025-08-07": { "cache_read_input_token_cost": 2.5e-08, @@ -27815,7 +27848,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27825,9 +27860,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5-nano": { "cache_read_input_token_cost": 5e-09, @@ -27860,7 +27893,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27870,17 +27905,15 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5-nano-2025-08-07": { "cache_read_input_token_cost": 5e-09, "cache_read_input_token_cost_flex": 2.5e-09, "deprecation_date": "2026-12-11", "input_cost_per_token": 5e-08, - "input_cost_per_token_priority": 2.5e-06, "input_cost_per_token_flex": 2.5e-08, + "input_cost_per_token_priority": 2.5e-06, "litellm_provider": "openai", "max_input_tokens": 272000, "max_output_tokens": 128000, @@ -27906,7 +27939,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, "supports_native_streaming": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -27916,9 +27951,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-image-1": { "cache_read_input_token_cost": 1.25e-06, @@ -28116,6 +28149,7 @@ "gpt-realtime-mini": { "cache_creation_input_audio_token_cost": 3e-07, "cache_read_input_audio_token_cost": 3e-07, + "cache_read_input_token_cost": 6e-08, "deprecation_date": "2027-01-20", "input_cost_per_audio_token": 1e-05, "input_cost_per_token": 6e-07, @@ -28143,8 +28177,7 @@ "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "cache_read_input_token_cost": 6e-08 + "supports_tool_choice": true }, "gpt-realtime-2025-08-28": { "cache_creation_input_audio_token_cost": 4e-07, @@ -28181,6 +28214,8 @@ }, "gradient_ai/alibaba-qwen3-32b": { "litellm_provider": "gradient_ai", + "max_input_tokens": 131072, + "max_output_tokens": 40960, "max_tokens": 40960, "mode": "chat", "supported_endpoints": [ @@ -28189,13 +28224,13 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 131072, - "max_output_tokens": 40960 + "supports_tool_choice": false }, "gradient_ai/anthropic-claude-3-opus": { "input_cost_per_token": 1.5e-05, "litellm_provider": "gradient_ai", + "max_input_tokens": 200000, + "max_output_tokens": 1024, "max_tokens": 1024, "mode": "chat", "output_cost_per_token": 7.5e-05, @@ -28205,13 +28240,13 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 200000, - "max_output_tokens": 1024 + "supports_tool_choice": false }, "gradient_ai/anthropic-claude-3.5-haiku": { "input_cost_per_token": 8e-07, "litellm_provider": "gradient_ai", + "max_input_tokens": 200000, + "max_output_tokens": 1024, "max_tokens": 1024, "mode": "chat", "output_cost_per_token": 4e-06, @@ -28221,13 +28256,13 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 200000, - "max_output_tokens": 1024 + "supports_tool_choice": false }, "gradient_ai/anthropic-claude-3.5-sonnet": { "input_cost_per_token": 3e-06, "litellm_provider": "gradient_ai", + "max_input_tokens": 200000, + "max_output_tokens": 1024, "max_tokens": 1024, "mode": "chat", "output_cost_per_token": 1.5e-05, @@ -28237,13 +28272,13 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 200000, - "max_output_tokens": 1024 + "supports_tool_choice": false }, "gradient_ai/anthropic-claude-3.7-sonnet": { "input_cost_per_token": 3e-06, "litellm_provider": "gradient_ai", + "max_input_tokens": 200000, + "max_output_tokens": 1024, "max_tokens": 1024, "mode": "chat", "output_cost_per_token": 1.5e-05, @@ -28253,13 +28288,13 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 200000, - "max_output_tokens": 1024 + "supports_tool_choice": false }, "gradient_ai/deepseek-r1-distill-llama-70b": { "input_cost_per_token": 9.9e-07, "litellm_provider": "gradient_ai", + "max_input_tokens": 32768, + "max_output_tokens": 8000, "max_tokens": 8000, "mode": "chat", "output_cost_per_token": 9.9e-07, @@ -28269,13 +28304,13 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 32768, - "max_output_tokens": 8000 + "supports_tool_choice": false }, "gradient_ai/llama3-8b-instruct": { "input_cost_per_token": 2e-07, "litellm_provider": "gradient_ai", + "max_input_tokens": 8192, + "max_output_tokens": 512, "max_tokens": 512, "mode": "chat", "output_cost_per_token": 2e-07, @@ -28285,13 +28320,13 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 8192, - "max_output_tokens": 512 + "supports_tool_choice": false }, "gradient_ai/llama3.3-70b-instruct": { "input_cost_per_token": 6.5e-07, "litellm_provider": "gradient_ai", + "max_input_tokens": 128000, + "max_output_tokens": 2048, "max_tokens": 2048, "mode": "chat", "output_cost_per_token": 6.5e-07, @@ -28301,13 +28336,13 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 128000, - "max_output_tokens": 2048 + "supports_tool_choice": false }, "gradient_ai/mistral-nemo-instruct-2407": { "input_cost_per_token": 3e-07, "litellm_provider": "gradient_ai", + "max_input_tokens": 128000, + "max_output_tokens": 512, "max_tokens": 512, "mode": "chat", "output_cost_per_token": 3e-07, @@ -28317,12 +28352,12 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 128000, - "max_output_tokens": 512 + "supports_tool_choice": false }, "gradient_ai/openai-gpt-4o": { "litellm_provider": "gradient_ai", + "max_input_tokens": 128000, + "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", "supported_endpoints": [ @@ -28331,12 +28366,12 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 128000, - "max_output_tokens": 16384 + "supports_tool_choice": false }, "gradient_ai/openai-gpt-4o-mini": { "litellm_provider": "gradient_ai", + "max_input_tokens": 128000, + "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", "supported_endpoints": [ @@ -28345,13 +28380,13 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 128000, - "max_output_tokens": 16384 + "supports_tool_choice": false }, "gradient_ai/openai-o3": { "input_cost_per_token": 2e-06, "litellm_provider": "gradient_ai", + "max_input_tokens": 200000, + "max_output_tokens": 100000, "max_tokens": 100000, "mode": "chat", "output_cost_per_token": 8e-06, @@ -28361,13 +28396,13 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 200000, - "max_output_tokens": 100000 + "supports_tool_choice": false }, "gradient_ai/openai-o3-mini": { "input_cost_per_token": 1.1e-06, "litellm_provider": "gradient_ai", + "max_input_tokens": 200000, + "max_output_tokens": 100000, "max_tokens": 100000, "mode": "chat", "output_cost_per_token": 4.4e-06, @@ -28377,16 +28412,14 @@ "supported_modalities": [ "text" ], - "supports_tool_choice": false, - "max_input_tokens": 200000, - "max_output_tokens": 100000 + "supports_tool_choice": false }, "lemonade/Qwen3-Coder-30B-A3B-Instruct-GGUF": { "input_cost_per_token": 0, "litellm_provider": "lemonade", - "max_tokens": 32768, "max_input_tokens": 262144, "max_output_tokens": 32768, + "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 0, "supports_function_calling": true, @@ -28396,9 +28429,9 @@ "lemonade/gpt-oss-20b-mxfp4-GGUF": { "input_cost_per_token": 0, "litellm_provider": "lemonade", - "max_tokens": 32768, "max_input_tokens": 131072, "max_output_tokens": 32768, + "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 0, "supports_function_calling": true, @@ -28408,9 +28441,9 @@ "lemonade/gpt-oss-120b-mxfp-GGUF": { "input_cost_per_token": 0, "litellm_provider": "lemonade", - "max_tokens": 32768, "max_input_tokens": 131072, "max_output_tokens": 32768, + "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 0, "supports_function_calling": true, @@ -28420,9 +28453,9 @@ "lemonade/Gemma-3-4b-it-GGUF": { "input_cost_per_token": 0, "litellm_provider": "lemonade", - "max_tokens": 8192, "max_input_tokens": 128000, "max_output_tokens": 8192, + "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 0, "supports_function_calling": true, @@ -28432,9 +28465,9 @@ "lemonade/Qwen3-4B-Instruct-2507-GGUF": { "input_cost_per_token": 0, "litellm_provider": "lemonade", - "max_tokens": 32768, "max_input_tokens": 262144, "max_output_tokens": 32768, + "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 0, "supports_function_calling": true, @@ -28592,15 +28625,15 @@ "supports_vision": true }, "groq/moonshotai/kimi-k2-instruct-0905": { + "cache_read_input_token_cost": 5e-07, "deprecation_date": "2026-04-15", "input_cost_per_token": 1e-06, - "output_cost_per_token": 3e-06, - "cache_read_input_token_cost": 5e-07, "litellm_provider": "groq", "max_input_tokens": 262144, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 3e-06, "supports_function_calling": true, "supports_response_schema": true, "supports_tool_choice": true @@ -28695,6 +28728,7 @@ "mode": "audio_speech" }, "groq/qwen/qwen3.6-27b": { + "cache_read_input_token_cost": 3e-07, "input_cost_per_token": 6e-07, "litellm_provider": "groq", "max_input_tokens": 131072, @@ -28707,8 +28741,7 @@ "supports_reasoning": true, "supports_response_schema": false, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-07 + "supports_vision": true }, "groq/qwen/qwen3-32b": { "deprecation_date": "2026-07-17", @@ -28759,43 +28792,43 @@ }, "heroku/claude-3-5-haiku": { "litellm_provider": "heroku", + "max_input_tokens": 200000, + "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "max_input_tokens": 200000, - "max_output_tokens": 8192 + "supports_tool_choice": true }, "heroku/claude-3-5-sonnet-latest": { "litellm_provider": "heroku", + "max_input_tokens": 200000, + "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "max_input_tokens": 200000, - "max_output_tokens": 8192 + "supports_tool_choice": true }, "heroku/claude-3-7-sonnet": { "litellm_provider": "heroku", + "max_input_tokens": 200000, + "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "max_input_tokens": 200000, - "max_output_tokens": 8192 + "supports_tool_choice": true }, "heroku/claude-4-sonnet": { "litellm_provider": "heroku", + "max_input_tokens": 200000, + "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "max_input_tokens": 200000, - "max_output_tokens": 8192 + "supports_tool_choice": true }, "high/1024-x-1024/gpt-image-1": { "deprecation_date": "2026-10-23", @@ -29167,20 +29200,22 @@ "jp.anthropic.claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 4.125e-06, "cache_creation_input_token_cost_above_1hr": 6.6e-06, + "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.32e-05, + "cache_creation_input_token_cost_above_200k_tokens": 8.25e-06, "cache_read_input_token_cost": 3.3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6.6e-07, "input_cost_per_token": 3.3e-06, "input_cost_per_token_above_200k_tokens": 6.6e-06, - "output_cost_per_token_above_200k_tokens": 2.475e-05, - "cache_creation_input_token_cost_above_200k_tokens": 8.25e-06, - "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.32e-05, - "cache_read_input_token_cost_above_200k_tokens": 6.6e-07, + "input_cost_per_token_batches": 1.65e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.65e-05, + "output_cost_per_token_above_200k_tokens": 2.475e-05, + "output_cost_per_token_batches": 8.25e-06, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -29189,45 +29224,43 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024, - "input_cost_per_token_batches": 1.65e-06, - "output_cost_per_token_batches": 8.25e-06 + "supports_tool_search": true, + "supports_vision": true }, "jp.anthropic.claude-haiku-4-5-20251001-v1:0": { "cache_creation_input_token_cost": 1.375e-06, "cache_creation_input_token_cost_above_1hr": 2.2e-06, "cache_read_input_token_cost": 1.1e-07, "input_cost_per_token": 1.1e-06, + "input_cost_per_token_batches": 5.5e-07, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5.5e-06, + "output_cost_per_token_batches": 2.75e-06, + "prompt_cache_min_tokens": 4096, "source": "https://aws.amazon.com/about-aws/whats-new/2025/10/claude-4-5-haiku-anthropic-amazon-bedrock", "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096, - "input_cost_per_token_batches": 5.5e-07, - "output_cost_per_token_batches": 2.75e-06 + "supports_tool_search": true, + "supports_vision": true }, "crusoe/deepseek-ai/DeepSeek-R1-0528": { "input_cost_per_token": 3e-06, @@ -30132,8 +30165,8 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.2e-06, - "supports_system_messages": true, - "supports_native_structured_output": true + "supports_native_structured_output": true, + "supports_system_messages": true }, "minimax.minimax-m2.1": { "input_cost_per_token": 3e-07, @@ -30143,23 +30176,23 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "minimax.minimax-m2.5": { "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "minimax/speech-02-hd": { "input_cost_per_character": 0.0001, @@ -30194,98 +30227,98 @@ ] }, "minimax/MiniMax-M2.1": { - "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, - "cache_read_input_token_cost": 3e-08, "cache_creation_input_token_cost": 3.75e-07, + "cache_read_input_token_cost": 3e-08, + "input_cost_per_token": 3e-07, "litellm_provider": "minimax", + "max_input_tokens": 1000000, + "max_output_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.2e-06, "supports_function_calling": true, - "supports_tool_choice": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_system_messages": true, - "max_input_tokens": 1000000, - "max_output_tokens": 8192 + "supports_tool_choice": true }, "minimax/MiniMax-M2.1-lightning": { - "input_cost_per_token": 3e-07, - "output_cost_per_token": 2.4e-06, - "cache_read_input_token_cost": 3e-08, "cache_creation_input_token_cost": 3.75e-07, + "cache_read_input_token_cost": 3e-08, + "input_cost_per_token": 3e-07, "litellm_provider": "minimax", + "max_input_tokens": 1000000, + "max_output_tokens": 8192, "mode": "chat", + "output_cost_per_token": 2.4e-06, "supports_function_calling": true, - "supports_tool_choice": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_system_messages": true, - "max_input_tokens": 1000000, - "max_output_tokens": 8192 + "supports_tool_choice": true }, "minimax/MiniMax-M2.5": { - "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, - "cache_read_input_token_cost": 3e-08, "cache_creation_input_token_cost": 3.75e-07, + "cache_read_input_token_cost": 3e-08, + "input_cost_per_token": 3e-07, "litellm_provider": "minimax", + "max_input_tokens": 1000000, + "max_output_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.2e-06, "supports_function_calling": true, - "supports_tool_choice": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_system_messages": true, - "max_input_tokens": 1000000, - "max_output_tokens": 8192 + "supports_tool_choice": true }, "minimax/MiniMax-M2.5-lightning": { - "input_cost_per_token": 3e-07, - "output_cost_per_token": 2.4e-06, - "cache_read_input_token_cost": 3e-08, "cache_creation_input_token_cost": 3.75e-07, + "cache_read_input_token_cost": 3e-08, + "input_cost_per_token": 3e-07, "litellm_provider": "minimax", + "max_input_tokens": 1000000, + "max_output_tokens": 8192, "mode": "chat", + "output_cost_per_token": 2.4e-06, "supports_function_calling": true, - "supports_tool_choice": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_system_messages": true, - "max_input_tokens": 1000000, - "max_output_tokens": 8192 + "supports_tool_choice": true }, "minimax/MiniMax-M2": { - "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, - "cache_read_input_token_cost": 3e-08, "cache_creation_input_token_cost": 3.75e-07, + "cache_read_input_token_cost": 3e-08, + "input_cost_per_token": 3e-07, "litellm_provider": "minimax", + "max_input_tokens": 200000, + "max_output_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.2e-06, "supports_function_calling": true, - "supports_tool_choice": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_system_messages": true, - "max_input_tokens": 200000, - "max_output_tokens": 8192 + "supports_tool_choice": true }, "minimax/MiniMax-M3": { - "input_cost_per_token": 3e-07, - "input_cost_per_token_above_512k_tokens": 6e-07, - "output_cost_per_token": 1.2e-06, - "output_cost_per_token_above_512k_tokens": 2.4e-06, "cache_read_input_token_cost": 6e-08, "cache_read_input_token_cost_above_512k_tokens": 1.2e-07, + "input_cost_per_token": 3e-07, + "input_cost_per_token_above_512k_tokens": 6e-07, "litellm_provider": "minimax", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1.2e-06, + "output_cost_per_token_above_512k_tokens": 2.4e-06, "supports_function_calling": true, - "supports_tool_choice": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_vision": true, - "max_input_tokens": 1000000, - "max_output_tokens": 128000, - "supports_video_input": true + "supports_tool_choice": true, + "supports_video_input": true, + "supports_vision": true }, "mistral.devstral-2-123b": { "input_cost_per_token": 4e-07, @@ -30295,10 +30328,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "mistral.magistral-small-2509": { "input_cost_per_token": 5e-07, @@ -30321,8 +30354,8 @@ "mode": "chat", "output_cost_per_token": 2e-07, "supports_function_calling": true, - "supports_system_messages": true, - "supports_native_structured_output": true + "supports_native_structured_output": true, + "supports_system_messages": true }, "mistral.ministral-3-3b-instruct": { "input_cost_per_token": 1e-07, @@ -30333,8 +30366,8 @@ "mode": "chat", "output_cost_per_token": 1e-07, "supports_function_calling": true, - "supports_system_messages": true, - "supports_native_structured_output": true + "supports_native_structured_output": true, + "supports_system_messages": true }, "mistral.ministral-3-8b-instruct": { "input_cost_per_token": 1.5e-07, @@ -30345,8 +30378,8 @@ "mode": "chat", "output_cost_per_token": 1.5e-07, "supports_function_calling": true, - "supports_system_messages": true, - "supports_native_structured_output": true + "supports_native_structured_output": true, + "supports_system_messages": true }, "mistral.mistral-7b-instruct-v0:2": { "input_cost_per_token": 1.5e-07, @@ -30388,8 +30421,8 @@ "mode": "chat", "output_cost_per_token": 1.5e-06, "supports_function_calling": true, - "supports_system_messages": true, - "supports_native_structured_output": true + "supports_native_structured_output": true, + "supports_system_messages": true }, "mistral.mistral-small-2402-v1:0": { "input_cost_per_token": 1e-06, @@ -30420,8 +30453,8 @@ "mode": "chat", "output_cost_per_token": 4e-08, "supports_audio_input": true, - "supports_system_messages": true, - "supports_native_structured_output": true + "supports_native_structured_output": true, + "supports_system_messages": true }, "mistral.voxtral-small-24b-2507": { "input_cost_per_token": 1e-07, @@ -30432,8 +30465,8 @@ "mode": "chat", "output_cost_per_token": 3e-07, "supports_audio_input": true, - "supports_system_messages": true, - "supports_native_structured_output": true + "supports_native_structured_output": true, + "supports_system_messages": true }, "mistral/codestral-2405": { "deprecation_date": "2025-06-16", @@ -30470,11 +30503,11 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 9e-07, - "supports_assistant_prefill": true, - "supports_response_schema": true, - "supports_tool_choice": true, "source": "https://docs.mistral.ai/models/model-cards/codestral-25-08", - "supports_function_calling": true + "supports_assistant_prefill": true, + "supports_function_calling": true, + "supports_response_schema": true, + "supports_tool_choice": true }, "mistral/codestral-mamba-latest": { "input_cost_per_token": 2.5e-07, @@ -30688,24 +30721,24 @@ "supports_tool_choice": true }, "mistral/mistral-ocr-latest": { - "litellm_provider": "mistral", - "ocr_cost_per_page": 0.004, "annotation_cost_per_page": 0.005, + "litellm_provider": "mistral", "mode": "ocr", + "ocr_cost_per_page": 0.004, + "source": "https://mistral.ai/pricing#api-pricing", "supported_endpoints": [ "/v1/ocr" - ], - "source": "https://mistral.ai/pricing#api-pricing" + ] }, "mistral/mistral-ocr-4-0": { - "litellm_provider": "mistral", - "ocr_cost_per_page": 0.004, "annotation_cost_per_page": 0.005, + "litellm_provider": "mistral", "mode": "ocr", + "ocr_cost_per_page": 0.004, + "source": "https://mistral.ai/pricing#api-pricing", "supported_endpoints": [ "/v1/ocr" - ], - "source": "https://mistral.ai/pricing#api-pricing" + ] }, "mistral/mistral-ocr-4-1": { "annotation_cost_per_page": 0.005, @@ -30718,25 +30751,25 @@ ] }, "mistral/mistral-ocr-2505-completion": { + "annotation_cost_per_page": 0.003, "deprecation_date": "2026-05-31", "litellm_provider": "mistral", - "ocr_cost_per_page": 0.001, - "annotation_cost_per_page": 0.003, "mode": "ocr", + "ocr_cost_per_page": 0.001, + "source": "https://mistral.ai/pricing#api-pricing", "supported_endpoints": [ "/v1/ocr" - ], - "source": "https://mistral.ai/pricing#api-pricing" + ] }, "mistral/mistral-ocr-2512": { - "litellm_provider": "mistral", - "ocr_cost_per_page": 0.002, "annotation_cost_per_page": 0.003, + "litellm_provider": "mistral", "mode": "ocr", + "ocr_cost_per_page": 0.002, + "source": "https://mistral.ai/pricing#api-pricing", "supported_endpoints": [ "/v1/ocr" - ], - "source": "https://mistral.ai/pricing#api-pricing" + ] }, "mistral/magistral-medium-latest": { "input_cost_per_token": 2e-06, @@ -31051,9 +31084,9 @@ "source": "https://docs.mistral.ai/models/model-cards/mistral-small-4-0-26-03", "supports_assistant_prefill": true, "supports_function_calling": true, + "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_reasoning": true, "supports_vision": true }, "mistral/mistral-small-3-2-2506": { @@ -31292,9 +31325,9 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 2.5e-06, + "supports_native_structured_output": true, "supports_reasoning": true, - "supports_system_messages": true, - "supports_native_structured_output": true + "supports_system_messages": true }, "moonshotai.kimi-k2.5": { "input_cost_per_token": 6e-07, @@ -31304,11 +31337,11 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 3e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_vision": true }, "moonshot/kimi-k2-0711-preview": { "cache_read_input_token_cost": 1.5e-07, @@ -31391,42 +31424,6 @@ "supports_video_input": true, "supports_vision": true }, - "moonshot/kimi-k2.7-code": { - "cache_read_input_token_cost": 1.9e-07, - "input_cost_per_token": 9.5e-07, - "litellm_provider": "moonshot", - "max_input_tokens": 262144, - "max_output_tokens": 262144, - "max_tokens": 262144, - "mode": "chat", - "output_cost_per_token": 4e-06, - "source": "https://platform.kimi.ai/docs/pricing/chat-k27-code", - "supports_function_calling": true, - "supports_prompt_caching": true, - "supports_reasoning": true, - "supports_response_schema": true, - "supports_tool_choice": true, - "supports_video_input": true, - "supports_vision": true - }, - "moonshot/kimi-k2.7-code-highspeed": { - "cache_read_input_token_cost": 3.8e-07, - "input_cost_per_token": 1.9e-06, - "litellm_provider": "moonshot", - "max_input_tokens": 262144, - "max_output_tokens": 262144, - "max_tokens": 262144, - "mode": "chat", - "output_cost_per_token": 8e-06, - "source": "https://platform.kimi.ai/docs/pricing/chat-k27-code", - "supports_function_calling": true, - "supports_prompt_caching": true, - "supports_reasoning": true, - "supports_response_schema": true, - "supports_tool_choice": true, - "supports_video_input": true, - "supports_vision": true - }, "moonshot/kimi-k3": { "cache_read_input_token_cost": 3e-07, "input_cost_per_token": 3e-06, @@ -31712,6 +31709,7 @@ "supports_vision": false }, "multimodalembedding": { + "deprecation_date": "2027-04-01", "input_cost_per_character": 2e-07, "input_cost_per_image": 0.0001, "input_cost_per_token": 8e-07, @@ -31732,8 +31730,7 @@ "text", "image", "video" - ], - "deprecation_date": "2027-04-01" + ] }, "multimodalembedding@001": { "deprecation_date": "2027-04-01", @@ -31905,332 +31902,332 @@ ] }, "nebius/deepseek-ai/DeepSeek-R1": { - "max_tokens": 128000, + "input_cost_per_token": 8e-07, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 8e-07, - "output_cost_per_token": 2.4e-06, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 2.4e-06, + "source": "https://nebius.com/prices-ai-studio", "supports_function_calling": true, - "supports_reasoning": true, - "source": "https://nebius.com/prices-ai-studio" + "supports_reasoning": true }, "nebius/deepseek-ai/DeepSeek-R1-0528": { - "max_tokens": 164000, + "input_cost_per_token": 8e-07, + "litellm_provider": "nebius", "max_input_tokens": 164000, "max_output_tokens": 164000, - "input_cost_per_token": 8e-07, - "output_cost_per_token": 2.4e-06, - "litellm_provider": "nebius", + "max_tokens": 164000, "mode": "chat", + "output_cost_per_token": 2.4e-06, + "source": "https://nebius.com/prices-ai-studio", "supports_function_calling": true, - "supports_reasoning": true, - "source": "https://nebius.com/prices-ai-studio" + "supports_reasoning": true }, "nebius/deepseek-ai/DeepSeek-R1-Distill-Llama-70B": { - "max_tokens": 128000, + "input_cost_per_token": 2.5e-07, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 2.5e-07, - "output_cost_per_token": 7.5e-07, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 7.5e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/deepseek-ai/DeepSeek-V3": { - "max_tokens": 128000, + "input_cost_per_token": 5e-07, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 1.5e-06, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 1.5e-06, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/deepseek-ai/DeepSeek-V3-0324": { - "max_tokens": 128000, + "input_cost_per_token": 5e-07, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 1.5e-06, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 1.5e-06, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/google/gemma-3-27b-it": { - "max_tokens": 128000, + "input_cost_per_token": 6e-08, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 6e-08, - "output_cost_per_token": 2e-07, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 2e-07, + "source": "https://nebius.com/prices-ai-studio", "supports_function_calling": true, - "supports_vision": true, - "source": "https://nebius.com/prices-ai-studio" + "supports_vision": true }, "nebius/meta-llama/Llama-3.3-70B-Instruct": { - "max_tokens": 128000, + "input_cost_per_token": 1.3e-07, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 1.3e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 4e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/meta-llama/Llama-Guard-3-8B": { - "max_tokens": 128000, + "input_cost_per_token": 2e-08, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 2e-08, - "output_cost_per_token": 6e-08, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 6e-08, "source": "https://nebius.com/prices-ai-studio" }, "nebius/meta-llama/Meta-Llama-3.1-8B-Instruct": { - "max_tokens": 128000, + "input_cost_per_token": 2e-08, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 2e-08, - "output_cost_per_token": 6e-08, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 6e-08, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/meta-llama/Meta-Llama-3.1-70B-Instruct": { - "max_tokens": 128000, + "input_cost_per_token": 1.3e-07, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 1.3e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 4e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/meta-llama/Meta-Llama-3.1-405B-Instruct": { - "max_tokens": 128000, + "input_cost_per_token": 1e-06, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 1e-06, - "output_cost_per_token": 3e-06, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 3e-06, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/mistralai/Mistral-Nemo-Instruct-2407": { - "max_tokens": 128000, + "input_cost_per_token": 4e-08, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 4e-08, - "output_cost_per_token": 1.2e-07, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 1.2e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/NousResearch/Hermes-3-Llama-3.1-405B": { - "max_tokens": 128000, + "input_cost_per_token": 1e-06, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 1e-06, - "output_cost_per_token": 3e-06, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 3e-06, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/nvidia/Llama-3.1-Nemotron-Ultra-253B-v1": { - "max_tokens": 128000, + "input_cost_per_token": 6e-07, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 6e-07, - "output_cost_per_token": 1.8e-06, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 1.8e-06, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/nvidia/Llama-3.3-Nemotron-Super-49B-v1": { - "max_tokens": 131072, + "input_cost_per_token": 1e-07, + "litellm_provider": "nebius", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "nebius", + "max_tokens": 131072, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 4e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/Qwen/Qwen3-235B-A22B": { - "max_tokens": 262144, + "input_cost_per_token": 2e-07, + "litellm_provider": "nebius", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "nebius", + "max_tokens": 262144, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 6e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/Qwen/Qwen3-32B": { - "max_tokens": 32768, + "input_cost_per_token": 1e-07, + "litellm_provider": "nebius", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 3e-07, - "litellm_provider": "nebius", + "max_tokens": 32768, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 3e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/Qwen/Qwen3-30B-A3B": { - "max_tokens": 32768, + "input_cost_per_token": 1e-07, + "litellm_provider": "nebius", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 3e-07, - "litellm_provider": "nebius", + "max_tokens": 32768, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 3e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/Qwen/Qwen3-14B": { - "max_tokens": 32768, + "input_cost_per_token": 8e-08, + "litellm_provider": "nebius", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 8e-08, - "output_cost_per_token": 2.4e-07, - "litellm_provider": "nebius", + "max_tokens": 32768, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 2.4e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/Qwen/Qwen3-4B": { - "max_tokens": 32768, + "input_cost_per_token": 8e-08, + "litellm_provider": "nebius", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 8e-08, - "output_cost_per_token": 2.4e-07, - "litellm_provider": "nebius", + "max_tokens": 32768, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 2.4e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/Qwen/QwQ-32B": { - "max_tokens": 32768, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "nebius", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 4.5e-07, - "litellm_provider": "nebius", + "max_tokens": 32768, "mode": "chat", + "output_cost_per_token": 4.5e-07, + "source": "https://nebius.com/prices-ai-studio", "supports_function_calling": true, - "supports_reasoning": true, - "source": "https://nebius.com/prices-ai-studio" + "supports_reasoning": true }, "nebius/Qwen/Qwen2.5-72B-Instruct": { - "max_tokens": 128000, + "input_cost_per_token": 1.3e-07, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 1.3e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 4e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/Qwen/Qwen2.5-32B-Instruct": { - "max_tokens": 128000, + "input_cost_per_token": 6e-08, + "litellm_provider": "nebius", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 6e-08, - "output_cost_per_token": 2e-07, - "litellm_provider": "nebius", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 2e-07, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/Qwen/Qwen2.5-Coder-7B": { - "max_tokens": 32768, + "input_cost_per_token": 1e-08, + "litellm_provider": "nebius", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-08, - "output_cost_per_token": 3e-08, - "litellm_provider": "nebius", + "max_tokens": 32768, "mode": "chat", - "supports_function_calling": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 3e-08, + "source": "https://nebius.com/prices-ai-studio", + "supports_function_calling": true }, "nebius/Qwen/Qwen2.5-VL-72B-Instruct": { - "max_tokens": 131072, + "input_cost_per_token": 1.3e-07, + "litellm_provider": "nebius", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1.3e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "nebius", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 4e-07, + "source": "https://nebius.com/prices-ai-studio", "supports_function_calling": true, - "supports_vision": true, - "source": "https://nebius.com/prices-ai-studio" + "supports_vision": true }, "nebius/Qwen/Qwen2-VL-72B-Instruct": { - "max_tokens": 131072, + "input_cost_per_token": 1.3e-07, + "litellm_provider": "nebius", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1.3e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "nebius", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 4e-07, + "source": "https://nebius.com/prices-ai-studio", "supports_function_calling": true, - "supports_vision": true, - "source": "https://nebius.com/prices-ai-studio" + "supports_vision": true }, "nebius/Qwen/Qwen2-VL-7B-Instruct": { - "max_tokens": 131072, + "input_cost_per_token": 2e-08, + "litellm_provider": "nebius", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-08, - "output_cost_per_token": 6e-08, - "litellm_provider": "nebius", + "max_tokens": 131072, "mode": "chat", - "supports_vision": true, - "source": "https://nebius.com/prices-ai-studio" + "output_cost_per_token": 6e-08, + "source": "https://nebius.com/prices-ai-studio", + "supports_vision": true }, "nebius/BAAI/bge-en-icl": { - "max_tokens": 32768, - "max_input_tokens": 32768, "input_cost_per_token": 1e-08, - "output_cost_per_token": 0.0, "litellm_provider": "nebius", + "max_input_tokens": 32768, + "max_tokens": 32768, "mode": "embedding", + "output_cost_per_token": 0.0, "source": "https://nebius.com/prices-ai-studio" }, "nebius/BAAI/bge-multilingual-gemma2": { - "max_tokens": 8192, - "max_input_tokens": 8192, "input_cost_per_token": 1e-08, - "output_cost_per_token": 0.0, "litellm_provider": "nebius", + "max_input_tokens": 8192, + "max_tokens": 8192, "mode": "embedding", + "output_cost_per_token": 0.0, "source": "https://nebius.com/prices-ai-studio" }, "nebius/intfloat/e5-mistral-7b-instruct": { - "max_tokens": 32768, - "max_input_tokens": 32768, "input_cost_per_token": 1e-08, - "output_cost_per_token": 0.0, "litellm_provider": "nebius", + "max_input_tokens": 32768, + "max_tokens": 32768, "mode": "embedding", + "output_cost_per_token": 0.0, "source": "https://nebius.com/prices-ai-studio" }, "nvidia.nemotron-nano-12b-v2": { @@ -32262,11 +32259,11 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 2.4e-07, - "supports_function_calling": true, - "supports_system_messages": true, - "supports_tool_choice": true, "source": "https://aws.amazon.com/bedrock/pricing/", - "supports_native_structured_output": true + "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "nvidia.nemotron-super-3-120b": { "input_cost_per_token": 1.5e-07, @@ -32816,8 +32813,8 @@ "output_cost_per_token": 7.2e-07, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/meta.llama-3.1-70b-instruct": { "input_cost_per_token": 7.2e-07, @@ -32829,8 +32826,8 @@ "output_cost_per_token": 7.2e-07, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/meta.llama-3.1-405b-instruct": { "input_cost_per_token": 1.068e-05, @@ -32842,8 +32839,8 @@ "output_cost_per_token": 1.068e-05, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/meta.llama-3.2-90b-vision-instruct": { "input_cost_per_token": 2e-06, @@ -32855,8 +32852,8 @@ "output_cost_per_token": 2e-06, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, "supports_native_streaming": true, + "supports_response_schema": false, "supports_vision": true }, "oci/meta.llama-3.3-70b-instruct": { @@ -32869,8 +32866,8 @@ "output_cost_per_token": 7.2e-07, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/meta.llama-4-maverick-17b-128e-instruct-fp8": { "input_cost_per_token": 7.2e-07, @@ -32882,8 +32879,8 @@ "output_cost_per_token": 7.2e-07, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, "supports_native_streaming": true, + "supports_response_schema": false, "supports_vision": true }, "oci/meta.llama-4-scout-17b-16e-instruct": { @@ -32896,8 +32893,8 @@ "output_cost_per_token": 7.2e-07, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/xai.grok-3": { "input_cost_per_token": 3e-06, @@ -32909,8 +32906,8 @@ "output_cost_per_token": 1.5e-05, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/xai.grok-3-fast": { "input_cost_per_token": 5e-06, @@ -32922,8 +32919,8 @@ "output_cost_per_token": 2.5e-05, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/xai.grok-3-mini": { "input_cost_per_token": 3e-07, @@ -32935,8 +32932,8 @@ "output_cost_per_token": 5e-07, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/xai.grok-3-mini-fast": { "input_cost_per_token": 6e-07, @@ -32948,8 +32945,8 @@ "output_cost_per_token": 4e-06, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/xai.grok-4": { "input_cost_per_token": 3e-06, @@ -32961,8 +32958,8 @@ "output_cost_per_token": 1.5e-05, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/cohere.command-latest": { "input_cost_per_token": 1.56e-06, @@ -32974,8 +32971,8 @@ "output_cost_per_token": 1.56e-06, "source": "https://www.oracle.com/cloud/ai/generative-ai/pricing/", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/cohere.command-a-03-2025": { "input_cost_per_token": 1.56e-06, @@ -32987,8 +32984,8 @@ "output_cost_per_token": 1.56e-06, "source": "https://www.oracle.com/cloud/ai/generative-ai/pricing/", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/cohere.command-plus-latest": { "input_cost_per_token": 1.56e-06, @@ -33000,8 +32997,8 @@ "output_cost_per_token": 1.56e-06, "source": "https://www.oracle.com/cloud/ai/generative-ai/pricing/", "supports_function_calling": true, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/google.gemini-2.5-flash": { "input_cost_per_token": 1.5e-07, @@ -33013,10 +33010,10 @@ "output_cost_per_token": 6e-07, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": true, - "supports_vision": true, + "supports_image_size": false, "supports_native_streaming": true, - "supports_image_size": false + "supports_response_schema": true, + "supports_vision": true }, "oci/google.gemini-2.5-pro": { "input_cost_per_token": 1.25e-06, @@ -33028,9 +33025,9 @@ "output_cost_per_token": 1e-05, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, + "supports_native_streaming": true, "supports_response_schema": true, - "supports_vision": true, - "supports_native_streaming": true + "supports_vision": true }, "oci/google.gemini-2.5-flash-lite": { "input_cost_per_token": 7.5e-08, @@ -33042,10 +33039,10 @@ "output_cost_per_token": 3e-07, "source": "https://www.oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing", "supports_function_calling": true, - "supports_response_schema": true, - "supports_vision": true, + "supports_image_size": false, "supports_native_streaming": true, - "supports_image_size": false + "supports_response_schema": true, + "supports_vision": true }, "oci/cohere.command-a-vision": { "input_cost_per_token": 1.56e-06, @@ -33057,8 +33054,8 @@ "output_cost_per_token": 1.56e-06, "source": "https://www.oracle.com/cloud/ai/generative-ai/pricing/", "supports_function_calling": true, - "supports_response_schema": false, "supports_native_streaming": true, + "supports_response_schema": false, "supports_vision": true }, "oci/cohere.command-a-reasoning": { @@ -33071,8 +33068,8 @@ "output_cost_per_token": 1.56e-06, "source": "https://www.oracle.com/cloud/ai/generative-ai/pricing/", "supports_function_calling": false, - "supports_response_schema": false, - "supports_native_streaming": true + "supports_native_streaming": true, + "supports_response_schema": false }, "oci/cohere.embed-multilingual-image-v3.0": { "input_cost_per_token": 1e-07, @@ -33704,14 +33701,14 @@ "input_cost_per_image": 0.0004, "input_cost_per_token": 2.5e-07, "litellm_provider": "openrouter", + "max_input_tokens": 200000, + "max_output_tokens": 4096, "max_tokens": 4096, "mode": "chat", "output_cost_per_token": 1.25e-06, "supports_function_calling": true, "supports_tool_choice": true, - "supports_vision": true, - "max_input_tokens": 200000, - "max_output_tokens": 4096 + "supports_vision": true }, "openrouter/anthropic/claude-3.5-sonnet": { "input_cost_per_token": 3e-06, @@ -33744,9 +33741,9 @@ "supports_vision": true }, "openrouter/anthropic/claude-opus-4": { - "input_cost_per_image": 0.0048, "cache_creation_input_token_cost": 1.875e-05, "cache_read_input_token_cost": 1.5e-06, + "input_cost_per_image": 0.0048, "input_cost_per_token": 1.5e-05, "litellm_provider": "openrouter", "max_input_tokens": 200000, @@ -33763,10 +33760,10 @@ "supports_vision": true }, "openrouter/anthropic/claude-opus-4.1": { - "input_cost_per_image": 0.0048, "cache_creation_input_token_cost": 1.875e-05, "cache_creation_input_token_cost_above_1hr": 3e-05, "cache_read_input_token_cost": 1.5e-06, + "input_cost_per_image": 0.0048, "input_cost_per_token": 1.5e-05, "litellm_provider": "openrouter", "max_input_tokens": 200000, @@ -33774,41 +33771,40 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "openrouter/anthropic/claude-sonnet-4": { - "input_cost_per_image": 0.0048, "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "input_cost_per_image": 0.0048, "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, "litellm_provider": "openrouter", "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "openrouter/anthropic/claude-sonnet-4.6": { - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, @@ -33822,16 +33818,18 @@ "mode": "chat", "output_cost_per_token": 1.5e-05, "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "source": "https://openrouter.ai/anthropic/claude-sonnet-4.6", + "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "openrouter/anthropic/claude-opus-4.5": { "cache_creation_input_token_cost": 6.25e-06, @@ -33843,18 +33841,17 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_output_config": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "openrouter/anthropic/claude-opus-4.6": { - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, @@ -33864,39 +33861,41 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, + "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "openrouter/anthropic/claude-sonnet-4.5": { - "input_cost_per_image": 0.0048, "cache_creation_input_token_cost": 3.75e-06, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "input_cost_per_image": 0.0048, "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "litellm_provider": "openrouter", "max_input_tokens": 1000000, "max_output_tokens": 1000000, "max_tokens": 1000000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "openrouter/anthropic/claude-haiku-4.5": { "cache_creation_input_token_cost": 1.25e-06, @@ -33908,17 +33907,16 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, + "prompt_cache_min_tokens": 4096, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "openrouter/anthropic/claude-opus-4.7": { - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, @@ -33928,22 +33926,21 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 2048, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "openrouter/anthropic/claude-opus-5": { - "prompt_cache_min_tokens": 512, - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, @@ -33953,15 +33950,17 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 512, "source": "https://openrouter.ai/anthropic/claude-opus-5", + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_max_reasoning_effort": true, "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true @@ -34084,12 +34083,12 @@ "mode": "chat", "output_cost_per_token": 3.96e-06, "source": "https://openrouter.ai/deepseek/deepseek-v4-pro", + "supports_assistant_prefill": true, "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_tool_choice": true, - "supports_assistant_prefill": true + "supports_tool_choice": true }, "openrouter/deepseek/deepseek-v4-pro-0813": { "input_cost_per_token": 1.32e-06, @@ -34135,11 +34134,11 @@ "output_cost_per_token": 2.5e-06, "supports_audio_output": true, "supports_function_calling": true, + "supports_image_size": false, "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_image_size": false + "supports_vision": true }, "openrouter/google/gemini-2.5-pro": { "input_cost_per_audio_token": 7e-07, @@ -34158,9 +34157,9 @@ "supports_vision": true }, "openrouter/google/gemini-3-pro-preview": { + "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "input_cost_per_token": 2e-06, "input_cost_per_token_above_200k_tokens": 4e-06, "input_cost_per_token_batches": 1e-06, @@ -34326,9 +34325,9 @@ "tpm": 800000 }, "openrouter/google/gemini-3.1-pro-preview": { + "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "cache_creation_input_token_cost_above_200k_tokens": 2.5e-07, "input_cost_per_token": 2e-06, "input_cost_per_token_above_200k_tokens": 4e-06, "litellm_provider": "openrouter", @@ -34369,22 +34368,22 @@ "openrouter/mancer/weaver": { "input_cost_per_token": 5.625e-06, "litellm_provider": "openrouter", + "max_input_tokens": 8000, + "max_output_tokens": 2000, "max_tokens": 2000, "mode": "chat", "output_cost_per_token": 5.625e-06, - "supports_tool_choice": true, - "max_input_tokens": 8000, - "max_output_tokens": 2000 + "supports_tool_choice": true }, "openrouter/meta-llama/llama-3-70b-instruct": { "input_cost_per_token": 5.9e-07, "litellm_provider": "openrouter", + "max_input_tokens": 8192, + "max_output_tokens": 8000, "max_tokens": 8000, "mode": "chat", "output_cost_per_token": 7.9e-07, - "supports_tool_choice": true, - "max_input_tokens": 8192, - "max_output_tokens": 8000 + "supports_tool_choice": true }, "openrouter/minimax/minimax-m2": { "input_cost_per_token": 2.55e-07, @@ -34472,52 +34471,52 @@ "openrouter/mistralai/mistral-7b-instruct": { "input_cost_per_token": 1.3e-07, "litellm_provider": "openrouter", + "max_input_tokens": 32768, + "max_output_tokens": 8191, "max_tokens": 8191, "mode": "chat", "output_cost_per_token": 1.3e-07, - "supports_tool_choice": true, - "max_input_tokens": 32768, - "max_output_tokens": 8191 + "supports_tool_choice": true }, "openrouter/mistralai/mistral-large": { "input_cost_per_token": 8e-06, "litellm_provider": "openrouter", + "max_input_tokens": 128000, + "max_output_tokens": 8191, "max_tokens": 8191, "mode": "chat", "output_cost_per_token": 2.4e-05, - "supports_tool_choice": true, - "max_input_tokens": 128000, - "max_output_tokens": 8191 + "supports_tool_choice": true }, "openrouter/mistralai/mistral-small-3.1-24b-instruct": { "input_cost_per_token": 1e-07, "litellm_provider": "openrouter", + "max_input_tokens": 131072, + "max_output_tokens": 131072, "max_tokens": 131072, "mode": "chat", "output_cost_per_token": 3e-07, - "supports_tool_choice": true, - "max_input_tokens": 131072, - "max_output_tokens": 131072 + "supports_tool_choice": true }, "openrouter/mistralai/mistral-small-3.2-24b-instruct": { "input_cost_per_token": 1e-07, "litellm_provider": "openrouter", + "max_input_tokens": 128000, + "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 3e-07, - "supports_tool_choice": true, - "max_input_tokens": 128000, - "max_output_tokens": 128000 + "supports_tool_choice": true }, "openrouter/mistralai/mixtral-8x22b-instruct": { "input_cost_per_token": 6.5e-07, "litellm_provider": "openrouter", + "max_input_tokens": 65536, + "max_output_tokens": 65536, "max_tokens": 65536, "mode": "chat", "output_cost_per_token": 6.5e-07, - "supports_tool_choice": true, - "max_input_tokens": 65536, - "max_output_tokens": 65536 + "supports_tool_choice": true }, "openrouter/moonshotai/kimi-k2.5": { "cache_read_input_token_cost": 1e-07, @@ -34548,32 +34547,32 @@ "openrouter/openai/gpt-3.5-turbo": { "input_cost_per_token": 1.5e-06, "litellm_provider": "openrouter", + "max_input_tokens": 16385, + "max_output_tokens": 4096, "max_tokens": 4096, "mode": "chat", "output_cost_per_token": 2e-06, - "supports_tool_choice": true, - "max_input_tokens": 16385, - "max_output_tokens": 4096 + "supports_tool_choice": true }, "openrouter/openai/gpt-3.5-turbo-16k": { "input_cost_per_token": 3e-06, "litellm_provider": "openrouter", + "max_input_tokens": 16385, + "max_output_tokens": 4096, "max_tokens": 4096, "mode": "chat", "output_cost_per_token": 4e-06, - "supports_tool_choice": true, - "max_input_tokens": 16385, - "max_output_tokens": 4096 + "supports_tool_choice": true }, "openrouter/openai/gpt-4": { "input_cost_per_token": 3e-05, "litellm_provider": "openrouter", + "max_input_tokens": 8191, + "max_output_tokens": 4096, "max_tokens": 4096, "mode": "chat", "output_cost_per_token": 6e-05, - "supports_tool_choice": true, - "max_input_tokens": 8191, - "max_output_tokens": 4096 + "supports_tool_choice": true }, "openrouter/openai/gpt-4.1": { "cache_read_input_token_cost": 5e-07, @@ -34790,8 +34789,8 @@ "supports_vision": true }, "openrouter/openai/gpt-5.2": { - "input_cost_per_image": 0, "cache_read_input_token_cost": 1.75e-07, + "input_cost_per_image": 0, "input_cost_per_token": 1.75e-06, "litellm_provider": "openrouter", "max_input_tokens": 272000, @@ -34806,8 +34805,8 @@ "supports_vision": true }, "openrouter/openai/gpt-5.2-chat": { - "input_cost_per_image": 0, "cache_read_input_token_cost": 1.75e-07, + "input_cost_per_image": 0, "input_cost_per_token": 1.75e-06, "litellm_provider": "openrouter", "max_input_tokens": 128000, @@ -34940,8 +34939,8 @@ "mode": "chat", "output_cost_per_token": 9.5e-07, "source": "https://openrouter.ai/qwen/qwen3-coder", - "supports_tool_choice": true, - "supports_function_calling": true + "supports_function_calling": true, + "supports_tool_choice": true }, "openrouter/qwen/qwen3-coder-plus": { "input_cost_per_token": 1e-06, @@ -35095,12 +35094,12 @@ "openrouter/undi95/remm-slerp-l2-13b": { "input_cost_per_token": 1.875e-06, "litellm_provider": "openrouter", + "max_input_tokens": 6144, + "max_output_tokens": 4096, "max_tokens": 4096, "mode": "chat", "output_cost_per_token": 1.875e-06, - "supports_tool_choice": true, - "max_input_tokens": 6144, - "max_output_tokens": 4096 + "supports_tool_choice": true }, "openrouter/x-ai/grok-4": { "input_cost_per_token": 3e-06, @@ -35145,89 +35144,89 @@ "supports_tool_choice": true }, "openrouter/xiaomi/mimo-v2-flash": { - "input_cost_per_token": 1e-07, - "output_cost_per_token": 3e-07, "cache_creation_input_token_cost": 0.0, "cache_read_input_token_cost": 1e-08, + "input_cost_per_token": 1e-07, "litellm_provider": "openrouter", "max_input_tokens": 262144, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 3e-07, "supports_function_calling": true, - "supports_tool_choice": true, + "supports_prompt_caching": true, "supports_reasoning": true, - "supports_vision": false, - "supports_prompt_caching": true + "supports_tool_choice": true, + "supports_vision": false }, "openrouter/xiaomi/mimo-v2.5-pro": { - "input_cost_per_token": 1e-06, - "output_cost_per_token": 3e-06, "cache_creation_input_token_cost": 0.0, "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 1e-06, "litellm_provider": "openrouter", "max_input_tokens": 1048576, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 3e-06, "supports_function_calling": true, - "supports_tool_choice": true, + "supports_prompt_caching": true, "supports_reasoning": true, - "supports_vision": false, "supports_response_schema": true, - "supports_prompt_caching": true + "supports_tool_choice": true, + "supports_vision": false }, "openrouter/xiaomi/mimo-v2.5": { - "input_cost_per_token": 4e-07, - "output_cost_per_token": 2e-06, "cache_creation_input_token_cost": 0.0, "cache_read_input_token_cost": 8e-08, + "input_cost_per_token": 4e-07, "litellm_provider": "openrouter", "max_input_tokens": 1048576, "max_output_tokens": 131072, "max_tokens": 131072, "mode": "chat", - "supports_function_calling": true, - "supports_tool_choice": true, - "supports_reasoning": true, - "supports_vision": true, + "output_cost_per_token": 2e-06, "supports_audio_input": true, - "supports_video_input": true, + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, "supports_response_schema": true, - "supports_prompt_caching": true + "supports_tool_choice": true, + "supports_video_input": true, + "supports_vision": true }, "openrouter/z-ai/glm-4.7": { - "input_cost_per_token": 4e-07, - "output_cost_per_token": 1.5e-06, "cache_creation_input_token_cost": 0.0, "cache_read_input_token_cost": 0.0, + "input_cost_per_token": 4e-07, "litellm_provider": "openrouter", "max_input_tokens": 202752, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", + "output_cost_per_token": 1.5e-06, + "supports_assistant_prefill": true, "supports_function_calling": true, - "supports_tool_choice": true, - "supports_reasoning": true, - "supports_vision": true, "supports_prompt_caching": false, - "supports_assistant_prefill": true + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": true }, "openrouter/z-ai/glm-4.7-flash": { - "input_cost_per_token": 7e-08, - "output_cost_per_token": 4e-07, "cache_creation_input_token_cost": 0.0, "cache_read_input_token_cost": 0.0, + "input_cost_per_token": 7e-08, "litellm_provider": "openrouter", "max_input_tokens": 200000, "max_output_tokens": 32000, "max_tokens": 32000, "mode": "chat", + "output_cost_per_token": 4e-07, "supports_function_calling": true, - "supports_tool_choice": true, + "supports_prompt_caching": false, "supports_reasoning": true, - "supports_vision": true, - "supports_prompt_caching": false + "supports_tool_choice": true, + "supports_vision": true }, "openrouter/z-ai/glm-5": { "input_cost_per_token": 8e-07, @@ -35243,15 +35242,15 @@ "supports_tool_choice": true }, "openrouter/z-ai/glm-5.1": { - "input_cost_per_token": 1.05e-06, - "output_cost_per_token": 3.5e-06, - "cache_read_input_token_cost": 5.25e-07, "cache_creation_input_token_cost": 0.0, + "cache_read_input_token_cost": 5.25e-07, + "input_cost_per_token": 1.05e-06, "litellm_provider": "openrouter", "max_input_tokens": 202752, "max_output_tokens": 65535, "max_tokens": 65535, "mode": "chat", + "output_cost_per_token": 3.5e-06, "source": "https://openrouter.ai/z-ai/glm-5.1", "supports_function_calling": true, "supports_prompt_caching": true, @@ -35259,74 +35258,74 @@ "supports_tool_choice": true }, "openrouter/minimax/minimax-m2.1": { - "input_cost_per_token": 2.7e-07, - "output_cost_per_token": 1.2e-06, "cache_creation_input_token_cost": 0.0, "cache_read_input_token_cost": 0.0, + "input_cost_per_token": 2.7e-07, "litellm_provider": "openrouter", "max_input_tokens": 204000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", + "output_cost_per_token": 1.2e-06, + "supports_computer_use": false, "supports_function_calling": true, - "supports_tool_choice": true, - "supports_reasoning": true, - "supports_vision": true, "supports_prompt_caching": false, - "supports_computer_use": false + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": true }, "openrouter/minimax/minimax-m2.5": { - "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.1e-06, "cache_read_input_token_cost": 1.5e-07, + "input_cost_per_token": 3e-07, "litellm_provider": "openrouter", "max_input_tokens": 196608, "max_output_tokens": 65536, "max_tokens": 65536, "mode": "chat", + "output_cost_per_token": 1.1e-06, "source": "https://openrouter.ai/minimax/minimax-m2.5", + "supports_computer_use": false, "supports_function_calling": true, - "supports_tool_choice": true, - "supports_reasoning": true, - "supports_vision": false, "supports_prompt_caching": true, - "supports_computer_use": false + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": false }, "openrouter/openrouter/auto": { "input_cost_per_token": 0, - "output_cost_per_token": 0, "litellm_provider": "openrouter", "max_input_tokens": 2000000, "max_tokens": 2000000, "mode": "chat", + "output_cost_per_token": 0, + "supports_audio_input": true, "supports_function_calling": true, - "supports_tool_choice": true, "supports_reasoning": true, "supports_response_schema": true, - "supports_vision": true, - "supports_audio_input": true, - "supports_video_input": true + "supports_tool_choice": true, + "supports_video_input": true, + "supports_vision": true }, "openrouter/openrouter/free": { "input_cost_per_token": 0, - "output_cost_per_token": 0, "litellm_provider": "openrouter", "max_input_tokens": 200000, "max_tokens": 200000, "mode": "chat", + "output_cost_per_token": 0, "supports_function_calling": true, - "supports_tool_choice": true, "supports_reasoning": true, "supports_response_schema": true, + "supports_tool_choice": true, "supports_vision": true }, "openrouter/openrouter/bodybuilder": { "input_cost_per_token": 0, - "output_cost_per_token": 0, "litellm_provider": "openrouter", "max_input_tokens": 128000, "max_tokens": 128000, - "mode": "chat" + "mode": "chat", + "output_cost_per_token": 0 }, "ovhcloud/DeepSeek-R1-Distill-Llama-70B": { "input_cost_per_token": 6.7e-07, @@ -35890,130 +35889,131 @@ "perplexity/preset/fast-search": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, - "supports_function_calling": true + "supports_function_calling": true, + "supports_web_search": true }, "perplexity/preset/pro-search": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, - "supports_function_calling": true + "supports_function_calling": true, + "supports_web_search": true }, "perplexity/preset/deep-research": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, - "supports_function_calling": true + "supports_function_calling": true, + "supports_web_search": true }, "perplexity/preset/advanced-deep-research": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, - "supports_function_calling": true + "supports_function_calling": true, + "supports_web_search": true }, "perplexity/openai/gpt-5.2": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": true, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/openai/gpt-5.1": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": false, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/openai/gpt-5-mini": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": false, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/anthropic/claude-opus-4-6": { - "supports_adaptive_thinking": true, "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, - "supports_reasoning": false, + "supports_adaptive_thinking": true, "supports_function_calling": true, - "supports_output_config": true + "supports_legacy_thinking": true, + "supports_output_config": true, + "supports_reasoning": false, + "supports_web_search": true }, "perplexity/anthropic/claude-opus-4-7": { - "supports_adaptive_thinking": true, "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, - "supports_reasoning": false, + "supports_adaptive_thinking": true, "supports_function_calling": true, - "supports_output_config": true + "supports_output_config": true, + "supports_reasoning": false, + "supports_web_search": true }, "perplexity/anthropic/claude-opus-4-5": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, - "supports_reasoning": false, "supports_function_calling": true, - "supports_output_config": true + "supports_output_config": true, + "supports_reasoning": false, + "supports_web_search": true }, "perplexity/anthropic/claude-sonnet-4-5": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": false, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/anthropic/claude-haiku-4-5": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": false, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/google/gemini-3-pro-preview": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": false, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/google/gemini-3-flash-preview": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": false, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/google/gemini-2.5-pro": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": false, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/google/gemini-2.5-flash": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, - "supports_reasoning": false, "supports_function_calling": true, - "supports_image_size": false + "supports_image_size": false, + "supports_reasoning": false, + "supports_web_search": true }, "perplexity/xai/grok-4-1-fast-non-reasoning": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": false, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/perplexity/sonar": { "litellm_provider": "perplexity", "mode": "responses", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": false, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/perplexity/deepseek-v4-flash-0731": { "cache_read_input_token_cost": 2.8e-08, @@ -36022,9 +36022,9 @@ "mode": "responses", "output_cost_per_token": 2.6e-07, "source": "https://docs.perplexity.ai/docs/agent-api/models", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": true, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/perplexity/glm-5.2": { "cache_read_input_token_cost": 1.4e-07, @@ -36033,9 +36033,9 @@ "mode": "responses", "output_cost_per_token": 4.4e-06, "source": "https://docs.perplexity.ai/docs/agent-api/models", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": true, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/perplexity/kimi-k3": { "cache_read_input_token_cost": 3e-07, @@ -36044,9 +36044,9 @@ "mode": "responses", "output_cost_per_token": 1.5e-05, "source": "https://docs.perplexity.ai/docs/agent-api/models", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": true, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/perplexity/kimi-k2.7-code": { "cache_read_input_token_cost": 1.9e-07, @@ -36055,9 +36055,9 @@ "mode": "responses", "output_cost_per_token": 4e-06, "source": "https://docs.perplexity.ai/docs/agent-api/models", - "supports_web_search": true, + "supports_function_calling": true, "supports_reasoning": false, - "supports_function_calling": true + "supports_web_search": true }, "perplexity/pplx-embed-v1-0.6b": { "input_cost_per_token": 4e-09, @@ -36101,8 +36101,8 @@ "output_cost_per_token": 0.0, "source": "https://platform.publicai.co/docs", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_reasoning": true + "supports_reasoning": true, + "supports_tool_choice": true }, "publicai/allenai/Olmo-3-32B-Think": { "input_cost_per_token": 0.0, @@ -36114,8 +36114,8 @@ "output_cost_per_token": 0.0, "source": "https://platform.publicai.co/docs", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_reasoning": true + "supports_reasoning": true, + "supports_tool_choice": true }, "qwen.qwen3-coder-480b-a35b-v1:0": { "input_cost_per_token": 2.2e-07, @@ -36126,24 +36126,24 @@ "mode": "chat", "output_cost_per_token": 1.8e-06, "supports_function_calling": true, + "supports_native_structured_output": true, "supports_reasoning": true, - "supports_tool_choice": true, - "supports_native_structured_output": true + "supports_tool_choice": true }, "qwen.qwen3-235b-a22b-2507-v1:0": { "input_cost_per_token": 2.2e-07, + "input_cost_per_token_batches": 1.1e-07, "litellm_provider": "bedrock_converse", "max_input_tokens": 262144, "max_output_tokens": 131072, "max_tokens": 131072, "mode": "chat", "output_cost_per_token": 8.8e-07, + "output_cost_per_token_batches": 4.4e-07, "supports_function_calling": true, - "supports_reasoning": true, - "supports_tool_choice": true, "supports_native_structured_output": true, - "input_cost_per_token_batches": 1.1e-07, - "output_cost_per_token_batches": 4.4e-07 + "supports_reasoning": true, + "supports_tool_choice": true }, "qwen.qwen3-coder-30b-a3b-v1:0": { "input_cost_per_token": 1.5e-07, @@ -36154,9 +36154,9 @@ "mode": "chat", "output_cost_per_token": 6e-07, "supports_function_calling": true, + "supports_native_structured_output": true, "supports_reasoning": true, - "supports_tool_choice": true, - "supports_native_structured_output": true + "supports_tool_choice": true }, "qwen.qwen3-32b-v1:0": { "input_cost_per_token": 1.5e-07, @@ -36167,9 +36167,9 @@ "mode": "chat", "output_cost_per_token": 6e-07, "supports_function_calling": true, + "supports_native_structured_output": true, "supports_reasoning": true, - "supports_tool_choice": true, - "supports_native_structured_output": true + "supports_tool_choice": true }, "qwen.qwen3-next-80b-a3b": { "input_cost_per_token": 1.5e-07, @@ -36180,8 +36180,8 @@ "mode": "chat", "output_cost_per_token": 1.2e-06, "supports_function_calling": true, - "supports_system_messages": true, - "supports_native_structured_output": true + "supports_native_structured_output": true, + "supports_system_messages": true }, "qwen.qwen3-vl-235b-a22b": { "input_cost_per_token": 5.3e-07, @@ -36192,9 +36192,9 @@ "mode": "chat", "output_cost_per_token": 2.66e-06, "supports_function_calling": true, + "supports_native_structured_output": true, "supports_system_messages": true, - "supports_vision": true, - "supports_native_structured_output": true + "supports_vision": true }, "qwen.qwen3-coder-next": { "input_cost_per_token": 5e-07, @@ -36204,10 +36204,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "reducto/parse-legacy": { "litellm_provider": "reducto", @@ -36377,300 +36377,301 @@ }, "replicate/openai/gpt-5": { "input_cost_per_token": 1.25e-06, - "output_cost_per_token": 1e-05, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 1e-05, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "replicate/openai/gpt-oss-20b": { "input_cost_per_token": 9e-08, - "output_cost_per_token": 3.6e-07, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 3.6e-07, "supports_function_calling": true, "supports_system_messages": true }, "replicate/anthropic/claude-4.5-haiku": { "input_cost_per_token": 1e-06, - "output_cost_per_token": 5e-06, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 5e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_prompt_caching": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true, - "supports_prompt_caching": true + "supports_vision": true }, "replicate/ibm-granite/granite-3.3-8b-instruct": { "input_cost_per_token": 3e-08, - "output_cost_per_token": 2.5e-07, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 2.5e-07, "supports_function_calling": true, "supports_system_messages": true }, "replicate/openai/gpt-4o": { "input_cost_per_token": 2.5e-06, - "output_cost_per_token": 1e-05, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 1e-05, + "supports_audio_input": true, + "supports_audio_output": true, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true, - "supports_audio_input": true, - "supports_audio_output": true + "supports_vision": true }, "replicate/openai/o4-mini": { "input_cost_per_token": 1e-06, - "output_cost_per_token": 4e-06, - "output_cost_per_reasoning_token": 4e-06, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_reasoning_token": 4e-06, + "output_cost_per_token": 4e-06, "supports_reasoning": true, "supports_system_messages": true }, "replicate/openai/o1-mini": { "input_cost_per_token": 1.1e-06, - "output_cost_per_token": 4.4e-06, - "output_cost_per_reasoning_token": 4.4e-06, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_reasoning_token": 4.4e-06, + "output_cost_per_token": 4.4e-06, "supports_reasoning": true, "supports_system_messages": true }, "replicate/openai/o1": { "input_cost_per_token": 1.5e-05, - "output_cost_per_token": 6e-05, - "output_cost_per_reasoning_token": 6e-05, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_reasoning_token": 6e-05, + "output_cost_per_token": 6e-05, "supports_reasoning": true, "supports_system_messages": true }, "replicate/openai/gpt-4o-mini": { "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 6e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "replicate/qwen/qwen3-235b-a22b-instruct-2507": { "input_cost_per_token": 2.64e-07, - "output_cost_per_token": 1.06e-06, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 1.06e-06, "supports_function_calling": true, "supports_system_messages": true }, "replicate/anthropic/claude-4-sonnet": { "input_cost_per_token": 3e-06, - "output_cost_per_token": 1.5e-05, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 1.5e-05, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_prompt_caching": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true, - "supports_prompt_caching": true + "supports_vision": true }, "replicate/deepseek-ai/deepseek-v3": { "input_cost_per_token": 1.45e-06, - "output_cost_per_token": 1.45e-06, "litellm_provider": "replicate", - "mode": "chat", "max_input_tokens": 65536, "max_output_tokens": 8192, "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 1.45e-06, "supports_function_calling": true, "supports_system_messages": true }, "replicate/anthropic/claude-3.7-sonnet": { "input_cost_per_token": 3e-06, - "output_cost_per_token": 1.5e-05, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 1.5e-05, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_prompt_caching": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true, - "supports_prompt_caching": true + "supports_vision": true }, "replicate/anthropic/claude-3.5-haiku": { "input_cost_per_token": 1e-06, - "output_cost_per_token": 5e-06, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 5e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_prompt_caching": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true, - "supports_prompt_caching": true + "supports_vision": true }, "replicate/anthropic/claude-3.5-sonnet": { "input_cost_per_token": 3.75e-06, - "output_cost_per_token": 1.875e-05, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 1.875e-05, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_prompt_caching": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true, - "supports_prompt_caching": true + "supports_vision": true }, "replicate/google/gemini-3-pro": { "input_cost_per_token": 2e-06, - "output_cost_per_token": 1.2e-05, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 1.2e-05, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "replicate/anthropic/claude-4.5-sonnet": { "input_cost_per_token": 3e-06, - "output_cost_per_token": 1.5e-05, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 1.5e-05, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_prompt_caching": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true, - "supports_prompt_caching": true + "supports_vision": true }, "replicate/openai/gpt-4.1": { "input_cost_per_token": 2e-06, - "output_cost_per_token": 8e-06, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 8e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "replicate/openai/gpt-4.1-nano": { "input_cost_per_token": 1e-07, - "output_cost_per_token": 4e-07, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 4e-07, "supports_function_calling": true, "supports_system_messages": true }, "replicate/openai/gpt-4.1-mini": { "input_cost_per_token": 4e-07, - "output_cost_per_token": 1.6e-06, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 1.6e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "replicate/openai/gpt-5-nano": { "input_cost_per_token": 5e-08, - "output_cost_per_token": 4e-07, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 4e-07, "supports_function_calling": true, "supports_system_messages": true }, "replicate/openai/gpt-5-mini": { "input_cost_per_token": 2.5e-07, - "output_cost_per_token": 2e-06, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 2e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "replicate/google/gemini-2.5-flash": { "input_cost_per_token": 2.5e-06, - "output_cost_per_token": 2.5e-06, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 2.5e-06, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_response_schema": true, - "supports_image_size": false + "supports_vision": true }, "replicate/openai/gpt-oss-120b": { "input_cost_per_token": 1.8e-07, - "output_cost_per_token": 7.2e-07, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 7.2e-07, "supports_function_calling": true, "supports_system_messages": true }, "replicate/deepseek-ai/deepseek-v3.1": { "input_cost_per_token": 6.72e-07, - "output_cost_per_token": 2.016e-06, "litellm_provider": "replicate", - "mode": "chat", "max_input_tokens": 163840, "max_output_tokens": 163840, "max_tokens": 163840, + "mode": "chat", + "output_cost_per_token": 2.016e-06, "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true }, "replicate/xai/grok-4": { "input_cost_per_token": 7.2e-06, - "output_cost_per_token": 3.6e-05, "litellm_provider": "replicate", "mode": "chat", + "output_cost_per_token": 3.6e-05, "supports_function_calling": true, "supports_system_messages": true }, "replicate/deepseek-ai/deepseek-r1": { "input_cost_per_token": 3.75e-06, - "output_cost_per_token": 1e-05, - "output_cost_per_reasoning_token": 1e-05, "litellm_provider": "replicate", - "mode": "chat", "max_input_tokens": 65536, "max_output_tokens": 8192, "max_tokens": 8192, + "mode": "chat", + "output_cost_per_reasoning_token": 1e-05, + "output_cost_per_token": 1e-05, "supports_reasoning": true, "supports_system_messages": true }, "rerank-english-v2.0": { + "deprecation_date": "2025-04-30", "input_cost_per_query": 0.002, "input_cost_per_token": 0.0, "litellm_provider": "cohere", @@ -36678,8 +36679,7 @@ "max_output_tokens": 4096, "max_tokens": 4096, "mode": "rerank", - "output_cost_per_token": 0.0, - "deprecation_date": "2025-04-30" + "output_cost_per_token": 0.0 }, "rerank-english-v3.0": { "input_cost_per_query": 0.002, @@ -36692,6 +36692,7 @@ "output_cost_per_token": 0.0 }, "rerank-multilingual-v2.0": { + "deprecation_date": "2025-04-30", "input_cost_per_query": 0.002, "input_cost_per_token": 0.0, "litellm_provider": "cohere", @@ -36699,8 +36700,7 @@ "max_output_tokens": 4096, "max_tokens": 4096, "mode": "rerank", - "output_cost_per_token": 0.0, - "deprecation_date": "2025-04-30" + "output_cost_per_token": 0.0 }, "rerank-multilingual-v3.0": { "input_cost_per_query": 0.002, @@ -36991,57 +36991,57 @@ "supports_tool_choice": true }, "sambanova/DeepSeek-V3.1": { - "max_tokens": 131072, + "input_cost_per_token": 3e-06, + "litellm_provider": "sambanova", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 3e-06, - "output_cost_per_token": 4.5e-06, - "litellm_provider": "sambanova", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 4.5e-06, + "source": "https://cloud.sambanova.ai/plans/pricing", "supports_function_calling": true, - "supports_tool_choice": true, "supports_reasoning": true, - "source": "https://cloud.sambanova.ai/plans/pricing" + "supports_tool_choice": true }, "sambanova/gpt-oss-120b": { - "max_tokens": 131072, + "input_cost_per_token": 2.2e-07, + "litellm_provider": "sambanova", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2.2e-07, - "output_cost_per_token": 5.9e-07, - "litellm_provider": "sambanova", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 5.9e-07, + "source": "https://cloud.sambanova.ai/plans/pricing", "supports_function_calling": true, - "supports_tool_choice": true, "supports_reasoning": true, - "source": "https://cloud.sambanova.ai/plans/pricing" + "supports_tool_choice": true }, "sambanova/DeepSeek-V3.2": { - "max_tokens": 32768, + "input_cost_per_token": 3e-06, + "litellm_provider": "sambanova", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 3e-06, - "output_cost_per_token": 4.5e-06, - "litellm_provider": "sambanova", + "max_tokens": 32768, "mode": "chat", - "supports_function_calling": true, - "supports_tool_choice": true, + "output_cost_per_token": 4.5e-06, "source": "https://cloud.sambanova.ai/plans/pricing", "supported_endpoints": [ "/v1/chat/completions" ], + "supports_function_calling": true, + "supports_tool_choice": true, "supports_vision": false }, "sambanova/gemma-4-31B-it": { - "max_tokens": 131072, + "input_cost_per_token": 3.8e-07, + "litellm_provider": "sambanova", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 3.8e-07, - "output_cost_per_token": 1.15e-06, - "litellm_provider": "sambanova", + "max_tokens": 131072, "mode": "chat", - "supports_vision": true, - "source": "https://cloud.sambanova.ai/plans/pricing" + "output_cost_per_token": 1.15e-06, + "source": "https://cloud.sambanova.ai/plans/pricing", + "supports_vision": true }, "scx-ai/GLM-5.2": { "cache_read_input_token_cost": 2.2e-07, @@ -37078,28 +37078,28 @@ "supports_vision": true }, "snowflake/claude-3-5-sonnet": { + "cache_read_input_token_cost": 3e-07, + "input_cost_per_token": 3e-06, "litellm_provider": "snowflake", "max_input_tokens": 200000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 3e-06, "output_cost_per_token": 1.5e-05, - "cache_read_input_token_cost": 3e-07, "supports_computer_use": true, "supports_function_calling": true, - "supports_vision": true, "supports_prompt_caching": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_vision": true }, "snowflake/deepseek-r1": { + "input_cost_per_token": 1.35e-06, "litellm_provider": "snowflake", "max_input_tokens": 128000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 1.35e-06, "output_cost_per_token": 5.4e-06, "supports_reasoning": true, "supports_system_messages": true @@ -37154,34 +37154,34 @@ "mode": "chat" }, "snowflake/llama3.1-405b": { + "input_cost_per_token": 1.2e-06, "litellm_provider": "snowflake", "max_input_tokens": 128000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 1.2e-06, "output_cost_per_token": 1.2e-06, "supports_function_calling": true, "supports_system_messages": true }, "snowflake/llama3.1-70b": { + "input_cost_per_token": 7.2e-07, "litellm_provider": "snowflake", "max_input_tokens": 128000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 7.2e-07, "output_cost_per_token": 7.2e-07, "supports_function_calling": true, "supports_system_messages": true }, "snowflake/llama3.1-8b": { + "input_cost_per_token": 2.4e-07, "litellm_provider": "snowflake", "max_input_tokens": 128000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 2.4e-07, "output_cost_per_token": 2.4e-07, "supports_system_messages": true }, @@ -37200,13 +37200,13 @@ "mode": "chat" }, "snowflake/llama3.3-70b": { - "max_tokens": 16384, + "input_cost_per_token": 7.2e-07, + "litellm_provider": "snowflake", "max_input_tokens": 128000, "max_output_tokens": 16384, - "input_cost_per_token": 7.2e-07, - "output_cost_per_token": 7.2e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 7.2e-07, "supports_function_calling": true, "supports_system_messages": true }, @@ -37225,16 +37225,16 @@ "mode": "chat" }, "snowflake/mistral-large2": { + "input_cost_per_token": 2e-06, "litellm_provider": "snowflake", "max_input_tokens": 128000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 2e-06, "output_cost_per_token": 6e-06, "supports_function_calling": true, - "supports_system_messages": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_system_messages": true }, "snowflake/mixtral-8x7b": { "litellm_provider": "snowflake", @@ -37272,13 +37272,13 @@ "mode": "chat" }, "snowflake/snowflake-llama-3.3-70b": { - "max_tokens": 16384, + "input_cost_per_token": 7.2e-07, + "litellm_provider": "snowflake", "max_input_tokens": 128000, "max_output_tokens": 16384, - "input_cost_per_token": 7.2e-07, - "output_cost_per_token": 7.2e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 7.2e-07, "supports_function_calling": true, "supports_system_messages": true }, @@ -37969,6 +37969,8 @@ "together_ai/deepseek-ai/DeepSeek-V3.1": { "input_cost_per_token": 6e-07, "litellm_provider": "together_ai", + "max_input_tokens": 128000, + "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", "output_cost_per_token": 1.7e-06, @@ -37976,9 +37978,7 @@ "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_reasoning": true, - "supports_tool_choice": true, - "max_input_tokens": 128000, - "max_output_tokens": 16384 + "supports_tool_choice": true }, "together_ai/meta-llama/Llama-3.2-3B-Instruct-Turbo": { "litellm_provider": "together_ai", @@ -38178,9 +38178,9 @@ "output_cost_per_token": 2.8e-06, "source": "https://www.together.ai/models/kimi-k2-5", "supports_function_calling": true, + "supports_reasoning": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_reasoning": true + "supports_vision": true }, "together_ai/moonshotai/Kimi-K2-Instruct-0905": { "input_cost_per_token": 1e-06, @@ -38249,37 +38249,37 @@ "input_cost_per_character": 4e-06, "litellm_provider": "aws_polly", "mode": "audio_speech", + "source": "https://aws.amazon.com/polly/pricing/", "supported_endpoints": [ "/v1/audio/speech" - ], - "source": "https://aws.amazon.com/polly/pricing/" + ] }, "aws_polly/neural": { "input_cost_per_character": 1.6e-05, "litellm_provider": "aws_polly", "mode": "audio_speech", + "source": "https://aws.amazon.com/polly/pricing/", "supported_endpoints": [ "/v1/audio/speech" - ], - "source": "https://aws.amazon.com/polly/pricing/" + ] }, "aws_polly/long-form": { "input_cost_per_character": 0.0001, "litellm_provider": "aws_polly", "mode": "audio_speech", + "source": "https://aws.amazon.com/polly/pricing/", "supported_endpoints": [ "/v1/audio/speech" - ], - "source": "https://aws.amazon.com/polly/pricing/" + ] }, "aws_polly/generative": { "input_cost_per_character": 3e-05, "litellm_provider": "aws_polly", "mode": "audio_speech", + "source": "https://aws.amazon.com/polly/pricing/", "supported_endpoints": [ "/v1/audio/speech" - ], - "source": "https://aws.amazon.com/polly/pricing/" + ] }, "us.amazon.nova-lite-v1:0": { "input_cost_per_token": 6e-08, @@ -38346,43 +38346,45 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 4e-06, + "prompt_cache_min_tokens": 2048, "supports_assistant_prefill": true, "supports_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_response_schema": true, - "supports_tool_choice": true, - "prompt_cache_min_tokens": 2048 + "supports_tool_choice": true }, "us.anthropic.claude-haiku-4-5-20251001-v1:0": { "cache_creation_input_token_cost": 1.375e-06, "cache_creation_input_token_cost_above_1hr": 2.2e-06, "cache_read_input_token_cost": 1.1e-07, "input_cost_per_token": 1.1e-06, + "input_cost_per_token_batches": 5.5e-07, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5.5e-06, + "output_cost_per_token_batches": 2.75e-06, + "prompt_cache_min_tokens": 4096, "source": "https://aws.amazon.com/about-aws/whats-new/2025/10/claude-4-5-haiku-anthropic-amazon-bedrock", "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096, - "input_cost_per_token_batches": 5.5e-07, - "output_cost_per_token_batches": 2.75e-06 + "supports_tool_search": true, + "supports_vision": true }, "us.anthropic.claude-3-5-sonnet-20240620-v1:0": { + "cache_creation_input_token_cost": 3.75e-06, + "cache_read_input_token_cost": 3e-07, "input_cost_per_token": 3e-06, "litellm_provider": "bedrock", "max_input_tokens": 200000, @@ -38394,9 +38396,7 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-07, - "cache_creation_input_token_cost": 3.75e-06 + "supports_vision": true }, "us.anthropic.claude-3-5-sonnet-20241022-v2:0": { "cache_creation_input_token_cost": 3.75e-06, @@ -38438,6 +38438,8 @@ "supports_vision": true }, "us.anthropic.claude-3-haiku-20240307-v1:0": { + "cache_creation_input_token_cost": 3.125e-07, + "cache_read_input_token_cost": 2.5e-08, "deprecation_date": "2026-09-10", "input_cost_per_token": 2.5e-07, "litellm_provider": "bedrock", @@ -38450,11 +38452,11 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 2.5e-08, - "cache_creation_input_token_cost": 3.125e-07 + "supports_vision": true }, "us.anthropic.claude-3-opus-20240229-v1:0": { + "cache_creation_input_token_cost": 1.875e-05, + "cache_read_input_token_cost": 1.5e-06, "input_cost_per_token": 1.5e-05, "litellm_provider": "bedrock", "max_input_tokens": 200000, @@ -38465,11 +38467,11 @@ "supports_function_calling": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 1.5e-06, - "cache_creation_input_token_cost": 1.875e-05 + "supports_vision": true }, "us.anthropic.claude-3-sonnet-20240229-v1:0": { + "cache_creation_input_token_cost": 3.75e-06, + "cache_read_input_token_cost": 3e-07, "deprecation_date": "2026-07-30", "input_cost_per_token": 3e-06, "litellm_provider": "bedrock", @@ -38482,9 +38484,7 @@ "supports_pdf_input": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "cache_read_input_token_cost": 3e-07, - "cache_creation_input_token_cost": 3.75e-06 + "supports_vision": true }, "us.anthropic.claude-opus-4-1-20250805-v1:0": { "cache_creation_input_token_cost": 1.875e-05, @@ -38497,6 +38497,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -38510,26 +38511,27 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "us.anthropic.claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 4.125e-06, "cache_creation_input_token_cost_above_1hr": 6.6e-06, + "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.32e-05, + "cache_creation_input_token_cost_above_200k_tokens": 8.25e-06, "cache_read_input_token_cost": 3.3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6.6e-07, "input_cost_per_token": 3.3e-06, "input_cost_per_token_above_200k_tokens": 6.6e-06, - "output_cost_per_token_above_200k_tokens": 2.475e-05, - "cache_creation_input_token_cost_above_200k_tokens": 8.25e-06, - "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.32e-05, - "cache_read_input_token_cost_above_200k_tokens": 6.6e-07, + "input_cost_per_token_batches": 1.65e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.65e-05, + "output_cost_per_token_above_200k_tokens": 2.475e-05, + "output_cost_per_token_batches": 8.25e-06, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -38538,74 +38540,72 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024, - "input_cost_per_token_batches": 1.65e-06, - "output_cost_per_token_batches": 8.25e-06 + "supports_tool_search": true, + "supports_vision": true }, "us-gov.anthropic.claude-sonnet-4-5-20250929-v1:0": { "cache_creation_input_token_cost": 4.5e-06, "cache_creation_input_token_cost_above_1hr": 7.2e-06, + "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.44e-05, + "cache_creation_input_token_cost_above_200k_tokens": 9e-06, "cache_read_input_token_cost": 3.6e-07, + "cache_read_input_token_cost_above_200k_tokens": 7.2e-07, "input_cost_per_token": 3.6e-06, "input_cost_per_token_above_200k_tokens": 7.2e-06, - "output_cost_per_token_above_200k_tokens": 2.7e-05, - "cache_creation_input_token_cost_above_200k_tokens": 9e-06, - "cache_creation_input_token_cost_above_1hr_above_200k_tokens": 1.44e-05, - "cache_read_input_token_cost_above_200k_tokens": 7.2e-07, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.8e-05, + "output_cost_per_token_above_200k_tokens": 2.7e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 1024 + "supports_tool_search": true, + "supports_vision": true }, "au.anthropic.claude-haiku-4-5-20251001-v1:0": { "cache_creation_input_token_cost": 1.375e-06, "cache_creation_input_token_cost_above_1hr": 2.2e-06, "cache_read_input_token_cost": 1.1e-07, "input_cost_per_token": 1.1e-06, + "input_cost_per_token_batches": 5.5e-07, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5.5e-06, + "output_cost_per_token_batches": 2.75e-06, + "prompt_cache_min_tokens": 4096, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096, - "input_cost_per_token_batches": 5.5e-07, - "output_cost_per_token_batches": 2.75e-06 + "supports_tool_search": true, + "supports_vision": true }, "us.anthropic.claude-opus-4-20250514-v1:0": { "cache_creation_input_token_cost": 1.875e-05, @@ -38617,6 +38617,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -38630,21 +38631,21 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "us.anthropic.claude-opus-4-5-20251101-v1:0": { + "bedrock_output_config_effort_ceiling": "high", "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_1hr": 1.1e-05, "cache_read_input_token_cost": 5.5e-07, "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.75e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -38653,30 +38654,30 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "high", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "global.anthropic.claude-opus-4-5-20251101-v1:0": { + "bedrock_output_config_effort_ceiling": "high", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -38685,29 +38686,29 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "high", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "eu.anthropic.claude-opus-4-5-20251101-v1:0": { + "bedrock_output_config_effort_ceiling": "high", "cache_creation_input_token_cost": 6.25e-06, "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, "litellm_provider": "bedrock_converse", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -38716,33 +38717,34 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_output_config": true, + "supports_parallel_tool_use_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_output_config": true, - "bedrock_output_config_effort_ceiling": "high", - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "us.anthropic.claude-sonnet-4-20250514-v1:0": { + "bedrock_converse_supports_strict_tools": false, "cache_creation_input_token_cost": 3.75e-06, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, "deprecation_date": "2026-10-14", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -38756,9 +38758,7 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "bedrock_converse_supports_strict_tools": false, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "us.deepseek.r1-v1:0": { "input_cost_per_token": 1.35e-06, @@ -38781,8 +38781,8 @@ "mode": "chat", "output_cost_per_token": 1.85e-06, "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, + "supports_reasoning": true, "supports_tool_choice": true }, "eu.deepseek.v3.2": { @@ -38794,8 +38794,8 @@ "mode": "chat", "output_cost_per_token": 2.22e-06, "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, + "supports_reasoning": true, "supports_tool_choice": true }, "us.meta.llama3-1-405b-instruct-v1:0": { @@ -39040,9 +39040,9 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 2.4e-07, - "supports_vision": true, "supports_function_calling": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_vision": true }, "vercel_ai_gateway/amazon/nova-micro": { "input_cost_per_token": 3.5e-08, @@ -39063,9 +39063,9 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 3.2e-06, - "supports_vision": true, "supports_function_calling": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_vision": true }, "vercel_ai_gateway/amazon/titan-embed-text-v2": { "input_cost_per_token": 2e-08, @@ -39086,10 +39086,10 @@ "max_tokens": 4096, "mode": "chat", "output_cost_per_token": 1.25e-06, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/anthropic/claude-3-opus": { "cache_creation_input_token_cost": 1.875e-05, @@ -39101,10 +39101,10 @@ "max_tokens": 4096, "mode": "chat", "output_cost_per_token": 7.5e-05, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/anthropic/claude-3.5-haiku": { "cache_creation_input_token_cost": 1e-06, @@ -39116,10 +39116,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 4e-06, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/anthropic/claude-3.5-sonnet": { "cache_creation_input_token_cost": 3.75e-06, @@ -39131,10 +39131,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.5e-05, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/anthropic/claude-3.7-sonnet": { "cache_creation_input_token_cost": 3.75e-06, @@ -39146,10 +39146,10 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/anthropic/claude-4-opus": { "cache_creation_input_token_cost": 1.875e-05, @@ -39161,10 +39161,10 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/anthropic/claude-4-sonnet": { "cache_creation_input_token_cost": 3.75e-06, @@ -39244,6 +39244,7 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, + "prompt_cache_min_tokens": 4096, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, @@ -39251,8 +39252,7 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "vercel_ai_gateway/anthropic/claude-opus-4": { "cache_creation_input_token_cost": 1.875e-05, @@ -39283,6 +39283,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, @@ -39290,8 +39291,7 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "vercel_ai_gateway/anthropic/claude-opus-4.5": { "cache_creation_input_token_cost": 6.25e-06, @@ -39303,19 +39303,18 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_output_config": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "vercel_ai_gateway/anthropic/claude-opus-4.6": { - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_read_input_token_cost": 5e-07, "input_cost_per_token": 5e-06, @@ -39325,16 +39324,18 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, + "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_output_config": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "vercel_ai_gateway/anthropic/claude-sonnet-4": { "cache_creation_input_token_cost": 3.75e-06, @@ -39346,6 +39347,7 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, @@ -39353,8 +39355,7 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "vercel_ai_gateway/anthropic/claude-sonnet-4.5": { "cache_creation_input_token_cost": 3.75e-06, @@ -39366,14 +39367,14 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "prompt_cache_min_tokens": 1024, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "vercel_ai_gateway/cohere/command-a": { "input_cost_per_token": 2.5e-06, @@ -39384,8 +39385,8 @@ "mode": "chat", "output_cost_per_token": 1e-05, "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "vercel_ai_gateway/cohere/command-r": { "input_cost_per_token": 1.5e-07, @@ -39437,8 +39438,8 @@ "mode": "chat", "output_cost_per_token": 9.9e-07, "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "vercel_ai_gateway/deepseek/deepseek-v3": { "input_cost_per_token": 9e-07, @@ -39459,10 +39460,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 6e-07, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/google/gemini-2.0-flash-lite": { "deprecation_date": "2026-06-01", @@ -39473,10 +39474,10 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 3e-07, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/google/gemini-2.5-flash": { "input_cost_per_token": 3e-07, @@ -39486,11 +39487,11 @@ "max_tokens": 65536, "mode": "chat", "output_cost_per_token": 2.5e-06, - "supports_vision": true, "supports_function_calling": true, - "supports_tool_choice": true, + "supports_image_size": false, "supports_response_schema": true, - "supports_image_size": false + "supports_tool_choice": true, + "supports_vision": true }, "vercel_ai_gateway/google/gemini-2.5-pro": { "input_cost_per_token": 2.5e-06, @@ -39500,10 +39501,10 @@ "max_tokens": 65536, "mode": "chat", "output_cost_per_token": 1e-05, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/google/gemini-embedding-001": { "input_cost_per_token": 1.5e-07, @@ -39522,9 +39523,9 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 2e-07, - "supports_vision": true, "supports_function_calling": true, - "supports_tool_choice": true + "supports_tool_choice": true, + "supports_vision": true }, "vercel_ai_gateway/google/text-embedding-005": { "input_cost_per_token": 2.5e-08, @@ -39602,9 +39603,9 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 1.6e-07, - "supports_vision": true, "supports_function_calling": true, - "supports_tool_choice": true + "supports_tool_choice": true, + "supports_vision": true }, "vercel_ai_gateway/meta/llama-3.2-1b": { "input_cost_per_token": 1e-07, @@ -39634,9 +39635,9 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 7.2e-07, - "supports_vision": true, "supports_function_calling": true, - "supports_tool_choice": true + "supports_tool_choice": true, + "supports_vision": true }, "vercel_ai_gateway/meta/llama-3.3-70b": { "input_cost_per_token": 7.2e-07, @@ -39667,9 +39668,9 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 3e-07, - "supports_vision": true, "supports_function_calling": true, - "supports_tool_choice": true + "supports_tool_choice": true, + "supports_vision": true }, "vercel_ai_gateway/mistral/codestral": { "input_cost_per_token": 3e-07, @@ -39700,8 +39701,8 @@ "mode": "chat", "output_cost_per_token": 2.8e-07, "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "vercel_ai_gateway/mistral/magistral-medium": { "input_cost_per_token": 2e-06, @@ -39712,8 +39713,8 @@ "mode": "chat", "output_cost_per_token": 5e-06, "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "vercel_ai_gateway/mistral/magistral-small": { "input_cost_per_token": 5e-07, @@ -39744,9 +39745,9 @@ "max_tokens": 4000, "mode": "chat", "output_cost_per_token": 1e-07, - "supports_vision": true, "supports_function_calling": true, - "supports_tool_choice": true + "supports_tool_choice": true, + "supports_vision": true }, "vercel_ai_gateway/mistral/mistral-embed": { "input_cost_per_token": 1e-07, @@ -39786,8 +39787,8 @@ "mode": "chat", "output_cost_per_token": 3e-07, "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "vercel_ai_gateway/mistral/mixtral-8x22b-instruct": { "input_cost_per_token": 1.2e-06, @@ -39807,10 +39808,10 @@ "max_tokens": 4000, "mode": "chat", "output_cost_per_token": 1.5e-07, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/mistral/pixtral-large": { "input_cost_per_token": 2e-06, @@ -39820,10 +39821,10 @@ "max_tokens": 4000, "mode": "chat", "output_cost_per_token": 6e-06, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/moonshotai/kimi-k2": { "input_cost_per_token": 5.5e-07, @@ -39882,9 +39883,9 @@ "max_tokens": 4096, "mode": "chat", "output_cost_per_token": 3e-05, - "supports_vision": true, "supports_function_calling": true, - "supports_tool_choice": true + "supports_tool_choice": true, + "supports_vision": true }, "vercel_ai_gateway/openai/gpt-4.1": { "cache_creation_input_token_cost": 0.0, @@ -39896,10 +39897,10 @@ "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 8e-06, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/openai/gpt-4.1-mini": { "cache_creation_input_token_cost": 0.0, @@ -39911,10 +39912,10 @@ "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 1.6e-06, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/openai/gpt-4.1-nano": { "cache_creation_input_token_cost": 0.0, @@ -39926,10 +39927,10 @@ "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 4e-07, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/openai/gpt-4o": { "cache_creation_input_token_cost": 0.0, @@ -39941,10 +39942,10 @@ "max_tokens": 16384, "mode": "chat", "output_cost_per_token": 1e-05, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/openai/gpt-4o-mini": { "cache_creation_input_token_cost": 0.0, @@ -39956,10 +39957,10 @@ "max_tokens": 16384, "mode": "chat", "output_cost_per_token": 6e-07, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/openai/o1": { "cache_creation_input_token_cost": 0.0, @@ -39971,10 +39972,10 @@ "max_tokens": 100000, "mode": "chat", "output_cost_per_token": 6e-05, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/openai/o3": { "cache_creation_input_token_cost": 0.0, @@ -39986,10 +39987,10 @@ "max_tokens": 100000, "mode": "chat", "output_cost_per_token": 8e-06, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/openai/o3-mini": { "cache_creation_input_token_cost": 0.0, @@ -40002,8 +40003,8 @@ "mode": "chat", "output_cost_per_token": 4.4e-06, "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_tool_choice": true }, "vercel_ai_gateway/openai/o4-mini": { "cache_creation_input_token_cost": 0.0, @@ -40015,10 +40016,10 @@ "max_tokens": 100000, "mode": "chat", "output_cost_per_token": 4.4e-06, - "supports_vision": true, "supports_function_calling": true, + "supports_response_schema": true, "supports_tool_choice": true, - "supports_response_schema": true + "supports_vision": true }, "vercel_ai_gateway/openai/text-embedding-3-large": { "input_cost_per_token": 1.3e-07, @@ -40091,9 +40092,9 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 1.5e-05, - "supports_vision": true, "supports_function_calling": true, - "supports_tool_choice": true + "supports_tool_choice": true, + "supports_vision": true }, "vercel_ai_gateway/vercel/v0-1.5-md": { "input_cost_per_token": 3e-06, @@ -40103,9 +40104,9 @@ "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 1.5e-05, - "supports_vision": true, "supports_function_calling": true, - "supports_tool_choice": true + "supports_tool_choice": true, + "supports_vision": true }, "vercel_ai_gateway/xai/grok-2": { "input_cost_per_token": 2e-06, @@ -40126,9 +40127,9 @@ "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 1e-05, - "supports_vision": true, "supports_function_calling": true, - "supports_tool_choice": true + "supports_tool_choice": true, + "supports_vision": true }, "vercel_ai_gateway/xai/grok-3": { "input_cost_per_token": 3e-06, @@ -40207,9 +40208,9 @@ "supports_tool_choice": true }, "vercel_ai_gateway/zai/glm-4.6": { - "litellm_provider": "vercel_ai_gateway", "cache_read_input_token_cost": 1.1e-07, "input_cost_per_token": 4.5e-07, + "litellm_provider": "vercel_ai_gateway", "max_input_tokens": 200000, "max_output_tokens": 200000, "max_tokens": 200000, @@ -40269,10 +40270,10 @@ "supports_tool_choice": true }, "vertex_ai/claude-haiku-4-5": { - "deprecation_date": "2026-10-15", "cache_creation_input_token_cost": 1.25e-06, "cache_creation_input_token_cost_above_1hr": 2e-06, "cache_read_input_token_cost": 1e-07, + "deprecation_date": "2026-10-15", "input_cost_per_token": 1e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 200000, @@ -40280,24 +40281,24 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, + "prompt_cache_min_tokens": 4096, "regional_endpoint_uplift_multiplier": 1.1, "source": "https://cloud.google.com/vertex-ai/generative-ai/docs/partner-models/claude/haiku-4-5", "supports_assistant_prefill": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_native_streaming": true, - "supports_vision": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "vertex_ai/claude-haiku-4-5@20251001": { - "deprecation_date": "2026-10-15", "cache_creation_input_token_cost": 1.25e-06, "cache_creation_input_token_cost_above_1hr": 2e-06, "cache_read_input_token_cost": 1e-07, + "deprecation_date": "2026-10-15", "input_cost_per_token": 1e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 200000, @@ -40305,18 +40306,18 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, + "prompt_cache_min_tokens": 4096, "regional_endpoint_uplift_multiplier": 1.1, "source": "https://cloud.google.com/vertex-ai/generative-ai/docs/partner-models/claude/haiku-4-5", "supports_assistant_prefill": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_native_streaming": true, - "supports_vision": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "vertex_ai/claude-3-5-sonnet": { "input_cost_per_token": 3e-06, @@ -40448,10 +40449,10 @@ "supports_vision": true }, "vertex_ai/claude-opus-4": { - "deprecation_date": "2026-05-14", "cache_creation_input_token_cost": 1.875e-05, "cache_creation_input_token_cost_above_1hr": 3e-05, "cache_read_input_token_cost": 1.5e-06, + "deprecation_date": "2026-05-14", "input_cost_per_token": 1.5e-05, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 200000, @@ -40459,6 +40460,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -40472,14 +40474,13 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "vertex_ai/claude-opus-4-1": { - "deprecation_date": "2026-08-05", "cache_creation_input_token_cost": 1.875e-05, "cache_creation_input_token_cost_above_1hr": 3e-05, "cache_read_input_token_cost": 1.5e-06, + "deprecation_date": "2026-08-05", "input_cost_per_token": 1.5e-05, "input_cost_per_token_batches": 7.5e-06, "litellm_provider": "vertex_ai-anthropic_models", @@ -40495,10 +40496,10 @@ "supports_vision": true }, "vertex_ai/claude-opus-4-1@20250805": { - "deprecation_date": "2026-08-05", "cache_creation_input_token_cost": 1.875e-05, "cache_creation_input_token_cost_above_1hr": 3e-05, "cache_read_input_token_cost": 1.5e-06, + "deprecation_date": "2026-08-05", "input_cost_per_token": 1.5e-05, "input_cost_per_token_batches": 7.5e-06, "litellm_provider": "vertex_ai-anthropic_models", @@ -40514,10 +40515,10 @@ "supports_vision": true }, "vertex_ai/claude-opus-4-5": { - "deprecation_date": "2026-11-24", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2026-11-24", "input_cost_per_token": 5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 200000, @@ -40525,6 +40526,7 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, @@ -40534,20 +40536,19 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "vertex_ai/claude-opus-4-5@20251101": { - "deprecation_date": "2026-11-24", "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2026-11-24", "input_cost_per_token": 5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 200000, @@ -40555,6 +40556,7 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, @@ -40564,23 +40566,20 @@ "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, - "supports_pdf_input": true, - "supports_prompt_caching": true, - "supports_reasoning": true, - "supports_response_schema": true, - "supports_tool_choice": true, - "supports_vision": true, "supports_native_streaming": true, "supports_output_config": true, - "prompt_cache_min_tokens": 4096 + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true }, "vertex_ai/claude-opus-4-6": { - "deprecation_date": "2027-02-05", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-02-05", "input_cost_per_token": 5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -40588,31 +40587,32 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "vertex_ai/claude-opus-4-6@default": { - "deprecation_date": "2027-02-05", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-02-05", "input_cost_per_token": 5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -40620,31 +40620,32 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 4096, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, + "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_output_config": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 4096 + "supports_vision": true }, "vertex_ai/claude-opus-4-7": { - "deprecation_date": "2027-04-16", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-04-16", "input_cost_per_token": 5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -40652,14 +40653,18 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 2048, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -40667,17 +40672,13 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "vertex_ai/claude-opus-4-7@default": { - "deprecation_date": "2027-04-16", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-04-16", "input_cost_per_token": 5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -40685,14 +40686,18 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 2048, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -40700,17 +40705,13 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 2048 + "supports_xhigh_reasoning_effort": true }, "vertex_ai/claude-fable-5": { - "deprecation_date": "2027-06-08", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 1.25e-05, "cache_creation_input_token_cost_above_1hr": 2e-05, "cache_read_input_token_cost": 1e-06, + "deprecation_date": "2027-06-08", "input_cost_per_token": 1e-05, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -40718,16 +40719,19 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 5e-05, + "prompt_cache_min_tokens": 512, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "thinking_always_on": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -40736,16 +40740,13 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 512 + "thinking_always_on": true }, "vertex_ai/claude-fable-5@default": { - "deprecation_date": "2027-06-08", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 1.25e-05, "cache_creation_input_token_cost_above_1hr": 2e-05, "cache_read_input_token_cost": 1e-06, + "deprecation_date": "2027-06-08", "input_cost_per_token": 1e-05, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -40753,16 +40754,19 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 5e-05, + "prompt_cache_min_tokens": 512, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, "supports_adaptive_thinking": true, - "thinking_always_on": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -40771,17 +40775,13 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 512 + "thinking_always_on": true }, "vertex_ai/claude-opus-5": { - "deprecation_date": "2027-01-24", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_mid_conversation_system": true, - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-01-24", "input_cost_per_token": 5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -40789,14 +40789,19 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 512, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -40804,18 +40809,13 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 512 + "supports_xhigh_reasoning_effort": true }, "vertex_ai/claude-opus-5@default": { - "deprecation_date": "2027-01-24", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_mid_conversation_system": true, - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-01-24", "input_cost_per_token": 5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -40823,14 +40823,19 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 512, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -40838,18 +40843,13 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 512 + "supports_xhigh_reasoning_effort": true }, "vertex_ai/claude-opus-4-8": { - "deprecation_date": "2027-05-28", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_mid_conversation_system": true, - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-05-28", "input_cost_per_token": 5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -40857,14 +40857,19 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 1024, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -40872,18 +40877,13 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "vertex_ai/claude-opus-4-8@default": { - "deprecation_date": "2027-05-28", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_mid_conversation_system": true, - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, + "deprecation_date": "2027-05-28", "input_cost_per_token": 5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -40891,14 +40891,19 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, + "prompt_cache_min_tokens": 1024, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -40906,20 +40911,17 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "vertex_ai/claude-sonnet-4-5": { - "deprecation_date": "2026-09-29", "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "deprecation_date": "2026-09-29", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "input_cost_per_token_batches": 1.5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 200000, @@ -40927,7 +40929,9 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, "output_cost_per_token_batches": 7.5e-06, + "prompt_cache_min_tokens": 1024, "regional_endpoint_uplift_multiplier": 1.1, "supports_assistant_prefill": true, "supports_computer_use": true, @@ -40937,16 +40941,13 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "vertex_ai/claude-sonnet-5": { - "deprecation_date": "2026-12-24", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 2.5e-06, "cache_creation_input_token_cost_above_1hr": 4e-06, "cache_read_input_token_cost": 2e-07, + "deprecation_date": "2026-12-24", "input_cost_per_token": 2e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -40954,6 +40955,8 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1e-05, + "prompt_cache_min_tokens": 1024, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -40963,6 +40966,8 @@ "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -40970,13 +40975,9 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "vertex_ai/claude-sonnet-4-6": { - "regional_endpoint_uplift_multiplier": 1.1, - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, "cache_read_input_token_cost": 3e-07, @@ -40987,34 +40988,36 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1.5e-05, - "supports_assistant_prefill": true, - "supports_computer_use": true, - "supports_function_calling": true, - "supports_pdf_input": true, - "supports_prompt_caching": true, - "supports_reasoning": true, - "supports_response_schema": true, - "supports_max_reasoning_effort": true, - "supports_tool_choice": true, - "supports_vision": true, + "prompt_cache_min_tokens": 1024, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, + "supports_assistant_prefill": true, + "supports_computer_use": true, + "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, "supports_output_config": true, - "prompt_cache_min_tokens": 1024 + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true }, "vertex_ai/claude-sonnet-4-5@20250929": { - "deprecation_date": "2026-09-29", "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "deprecation_date": "2026-09-29", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "input_cost_per_token_batches": 1.5e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 200000, @@ -41022,25 +41025,26 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, "output_cost_per_token_batches": 7.5e-06, + "prompt_cache_min_tokens": 1024, "regional_endpoint_uplift_multiplier": 1.1, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_streaming": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "vertex_ai/claude-opus-4@20250514": { - "deprecation_date": "2026-05-14", "cache_creation_input_token_cost": 1.875e-05, "cache_creation_input_token_cost_above_1hr": 3e-05, "cache_read_input_token_cost": 1.5e-06, + "deprecation_date": "2026-05-14", "input_cost_per_token": 1.5e-05, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 200000, @@ -41048,6 +41052,7 @@ "max_tokens": 32000, "mode": "chat", "output_cost_per_token": 7.5e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -41061,25 +41066,25 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "vertex_ai/claude-sonnet-4": { - "deprecation_date": "2026-05-14", "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "deprecation_date": "2026-05-14", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -41093,25 +41098,25 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "vertex_ai/claude-sonnet-4@20250514": { - "deprecation_date": "2026-05-14", "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, "cache_read_input_token_cost": 3e-07, + "cache_read_input_token_cost_above_200k_tokens": 6e-07, + "deprecation_date": "2026-05-14", "input_cost_per_token": 3e-06, "input_cost_per_token_above_200k_tokens": 6e-06, - "output_cost_per_token_above_200k_tokens": 2.25e-05, - "cache_creation_input_token_cost_above_200k_tokens": 7.5e-06, - "cache_read_input_token_cost_above_200k_tokens": 6e-07, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, + "output_cost_per_token_above_200k_tokens": 2.25e-05, + "prompt_cache_min_tokens": 1024, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -41125,8 +41130,7 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "prompt_cache_min_tokens": 1024 + "supports_vision": true }, "vertex_ai/mistralai/codestral-2@001": { "input_cost_per_token": 3e-07, @@ -41262,8 +41266,8 @@ "supports_tool_choice": true }, "vertex_ai/gemini-2.5-flash-image": { - "deprecation_date": "2026-10-02", "cache_read_input_token_cost": 3e-08, + "deprecation_date": "2026-10-02", "input_cost_per_audio_token": 1e-06, "input_cost_per_token": 3e-07, "litellm_provider": "vertex_ai-language-models", @@ -41294,6 +41298,7 @@ ], "supports_audio_output": false, "supports_function_calling": true, + "supports_image_size": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -41304,8 +41309,7 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": false, - "tpm": 8000000, - "supports_image_size": false + "tpm": 8000000 }, "vertex_ai/gemini-3-pro-image": { "deprecation_date": "2027-05-28", @@ -41321,8 +41325,8 @@ "output_cost_per_image_token": 0.00012, "output_cost_per_token": 1.2e-05, "output_cost_per_token_batches": 6e-06, - "supports_reasoning": false, - "source": "https://docs.cloud.google.com/vertex-ai/generative-ai/docs/models/gemini/3-pro-image" + "source": "https://docs.cloud.google.com/vertex-ai/generative-ai/docs/models/gemini/3-pro-image", + "supports_reasoning": false }, "vertex_ai/gemini-3-pro-image-preview": { "input_cost_per_image": 0.0011, @@ -41337,8 +41341,8 @@ "output_cost_per_image_token": 0.00012, "output_cost_per_token": 1.2e-05, "output_cost_per_token_batches": 6e-06, - "supports_reasoning": false, - "source": "https://docs.cloud.google.com/vertex-ai/generative-ai/docs/models/gemini/3-pro-image" + "source": "https://docs.cloud.google.com/vertex-ai/generative-ai/docs/models/gemini/3-pro-image", + "supports_reasoning": false }, "vertex_ai/gemini-3.1-flash-image": { "deprecation_date": "2027-05-28", @@ -41352,8 +41356,8 @@ "output_cost_per_image": 0.0672, "output_cost_per_image_token": 6e-05, "output_cost_per_token": 3e-06, - "supports_reasoning": false, - "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models" + "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models", + "supports_reasoning": false }, "vertex_ai/gemini-3.1-flash-image-preview": { "input_cost_per_image": 0.00056, @@ -41366,8 +41370,8 @@ "output_cost_per_image": 0.0672, "output_cost_per_image_token": 6e-05, "output_cost_per_token": 3e-06, - "supports_reasoning": false, - "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models" + "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models", + "supports_reasoning": false }, "vertex_ai/gemini-3.1-flash-lite-image": { "cache_read_input_token_cost": 2.5e-08, @@ -41418,6 +41422,11 @@ "mode": "chat", "output_cost_per_reasoning_token": 1.5e-06, "output_cost_per_token": 1.5e-06, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models", "supported_endpoints": [ "/v1/chat/completions", @@ -41436,6 +41445,7 @@ "supports_audio_input": true, "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -41447,28 +41457,27 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "vertex_ai/gemini-3.1-flash-lite": { - "deprecation_date": "2027-05-07", "cache_read_input_token_cost": 2.5e-08, "cache_read_input_token_cost_flex": 1.25e-08, + "cache_read_input_token_cost_per_audio_token": 5e-08, "cache_read_input_token_cost_priority": 4.5e-08, + "deprecation_date": "2027-05-07", "input_cost_per_audio_token": 5e-07, "input_cost_per_token": 2.5e-07, "input_cost_per_token_batches": 1.25e-07, "input_cost_per_token_flex": 1.25e-07, "input_cost_per_token_priority": 4.5e-07, "litellm_provider": "vertex_ai-language-models", + "max_audio_length_hours": 8.4, + "max_audio_per_prompt": 1, + "max_images_per_prompt": 3000, "max_input_tokens": 1048576, "max_output_tokens": 65536, "max_tokens": 65536, + "max_videos_per_prompt": 10, "mode": "chat", "output_cost_per_reasoning_token": 1.5e-06, "output_cost_per_token": 1.5e-06, @@ -41476,6 +41485,11 @@ "output_cost_per_token_flex": 7.5e-07, "output_cost_per_token_priority": 2.7e-06, "regional_endpoint_uplift_multiplier": 1.1, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models", "supported_endpoints": [ "/v1/chat/completions", @@ -41493,39 +41507,29 @@ ], "supports_audio_input": true, "supports_audio_output": false, + "supports_code_execution": true, + "supports_file_search": true, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, + "supports_service_tier": true, "supports_system_messages": true, "supports_tool_choice": true, "supports_url_context": true, "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, - "web_search_billing_unit": "per_query", - "cache_read_input_token_cost_per_audio_token": 5e-08, - "max_audio_length_hours": 8.4, - "max_audio_per_prompt": 1, - "max_images_per_prompt": 3000, - "max_videos_per_prompt": 10, - "supports_code_execution": true, - "supports_file_search": true, - "supports_service_tier": true + "web_search_billing_unit": "per_query" }, "vertex_ai/gemini-3.5-flash-lite": { - "deprecation_date": "2027-07-21", "cache_read_input_token_cost": 3e-08, "cache_read_input_token_cost_flex": 2e-08, "cache_read_input_token_cost_priority": 5e-08, + "deprecation_date": "2027-07-21", "input_cost_per_token": 3e-07, "input_cost_per_token_batches": 1.5e-07, "input_cost_per_token_flex": 1.5e-07, @@ -41541,6 +41545,11 @@ "output_cost_per_token_flex": 1.25e-06, "output_cost_per_token_priority": 4.5e-06, "regional_endpoint_uplift_multiplier": 1.1, + "search_context_cost_per_query": { + "search_context_size_low": 0.014, + "search_context_size_medium": 0.014, + "search_context_size_high": 0.014 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models", "supported_endpoints": [ "/v1/chat/completions", @@ -41559,6 +41568,7 @@ "supports_audio_input": true, "supports_audio_output": false, "supports_function_calling": true, + "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -41570,12 +41580,6 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "supports_native_streaming": true, - "search_context_cost_per_query": { - "search_context_size_low": 0.014, - "search_context_size_medium": 0.014, - "search_context_size_high": 0.014 - }, "web_search_billing_unit": "per_query" }, "vertex_ai/deep-research-pro-preview-12-2025": { @@ -42059,17 +42063,17 @@ "litellm_provider": "vertex_ai", "mode": "ocr", "ocr_cost_per_page": 0.0005, + "source": "https://cloud.google.com/generative-ai-app-builder/pricing", "supported_endpoints": [ "/v1/ocr" - ], - "source": "https://cloud.google.com/generative-ai-app-builder/pricing" + ] }, "vertex_ai/deepseek-ai/deepseek-ocr-maas": { + "input_cost_per_token": 3e-07, "litellm_provider": "vertex_ai", "mode": "ocr", - "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, "ocr_cost_per_page": 0.0003, + "output_cost_per_token": 1.2e-06, "source": "https://cloud.google.com/vertex-ai/pricing", "supported_regions": [ "us-central1" @@ -42500,55 +42504,55 @@ "output_cost_per_token": 0.0 }, "wandb/openai/gpt-oss-120b": { - "max_tokens": 131000, + "input_cost_per_token": 3e-08, + "litellm_provider": "wandb", "max_input_tokens": 131000, "max_output_tokens": 131000, - "input_cost_per_token": 3e-08, - "output_cost_per_token": 1.7e-07, - "litellm_provider": "wandb", + "max_tokens": 131000, "mode": "chat", + "output_cost_per_token": 1.7e-07, "source": "https://wandb.ai/site/pricing/tokens/" }, "wandb/openai/gpt-oss-20b": { - "max_tokens": 131000, + "input_cost_per_token": 3e-08, + "litellm_provider": "wandb", "max_input_tokens": 131000, "max_output_tokens": 131000, - "input_cost_per_token": 3e-08, - "output_cost_per_token": 1.3e-07, - "litellm_provider": "wandb", + "max_tokens": 131000, "mode": "chat", + "output_cost_per_token": 1.3e-07, "source": "https://wandb.ai/site/pricing/tokens/" }, "wandb/Qwen/Qwen3-Coder-480B-A35B-Instruct": { - "max_tokens": 262000, + "deprecation_date": "2026-08-25", + "input_cost_per_token": 1e-06, + "litellm_provider": "wandb", "max_input_tokens": 262000, "max_output_tokens": 262000, - "input_cost_per_token": 1e-06, - "output_cost_per_token": 1.5e-06, - "litellm_provider": "wandb", + "max_tokens": 262000, "mode": "chat", - "deprecation_date": "2026-08-25", + "output_cost_per_token": 1.5e-06, "source": "https://wandb.ai/site/pricing/tokens/" }, "wandb/moonshotai/Kimi-K2-Instruct": { - "max_tokens": 128000, + "deprecation_date": "2026-03-04", + "input_cost_per_token": 6e-07, + "litellm_provider": "wandb", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 6e-07, - "output_cost_per_token": 2.5e-06, - "litellm_provider": "wandb", + "max_tokens": 128000, "mode": "chat", - "deprecation_date": "2026-03-04" + "output_cost_per_token": 2.5e-06 }, "wandb/moonshotai/Kimi-K2.5": { - "max_tokens": 262144, - "max_input_tokens": 262144, - "max_output_tokens": 262144, "cache_read_input_token_cost": 1e-07, "input_cost_per_token": 6e-07, - "output_cost_per_token": 3e-06, "litellm_provider": "wandb", + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "max_tokens": 262144, "mode": "chat", + "output_cost_per_token": 3e-06, "source": "https://wandb.ai/inference/coreweave/cw_moonshotai_Kimi-K2.5", "supports_function_calling": true, "supports_reasoning": true, @@ -42556,47 +42560,47 @@ "supports_vision": true }, "wandb/MiniMaxAI/MiniMax-M2.5": { - "max_tokens": 197000, + "deprecation_date": "2026-08-25", + "input_cost_per_token": 3e-07, + "litellm_provider": "wandb", "max_input_tokens": 197000, "max_output_tokens": 197000, - "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "wandb", + "max_tokens": 197000, "mode": "chat", + "output_cost_per_token": 1.2e-06, "source": "https://wandb.ai/site/pricing/tokens/", "supports_function_calling": true, "supports_reasoning": true, - "supports_response_schema": true, - "deprecation_date": "2026-08-25" + "supports_response_schema": true }, "wandb/meta-llama/Llama-3.1-8B-Instruct": { - "max_tokens": 128000, + "input_cost_per_token": 2.2e-07, + "litellm_provider": "wandb", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 2.2e-07, - "output_cost_per_token": 2.2e-07, - "litellm_provider": "wandb", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 2.2e-07, "source": "https://wandb.ai/site/pricing/tokens/" }, "wandb/deepseek-ai/DeepSeek-V3.1": { - "max_tokens": 161000, + "input_cost_per_token": 5.5e-07, + "litellm_provider": "wandb", "max_input_tokens": 161000, "max_output_tokens": 161000, - "input_cost_per_token": 5.5e-07, - "output_cost_per_token": 1.65e-06, - "litellm_provider": "wandb", + "max_tokens": 161000, "mode": "chat", + "output_cost_per_token": 1.65e-06, "source": "https://wandb.ai/site/pricing/tokens/" }, "wandb/meta-llama/Llama-3.3-70B-Instruct": { - "max_tokens": 128000, + "input_cost_per_token": 7.1e-07, + "litellm_provider": "wandb", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 7.1e-07, - "output_cost_per_token": 7.1e-07, - "litellm_provider": "wandb", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 7.1e-07, "source": "https://wandb.ai/site/pricing/tokens/" }, "watsonx/ibm/granite-3-8b-instruct": { @@ -42636,322 +42640,322 @@ "supports_vision": false }, "watsonx/bigscience/mt0-xxl-13b": { - "max_tokens": 8192, + "input_cost_per_token": 0.0005, + "litellm_provider": "watsonx", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 0.0005, - "output_cost_per_token": 0.002, - "litellm_provider": "watsonx", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 0.002, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/core42/jais-13b-chat": { - "max_tokens": 8192, + "input_cost_per_token": 0.0005, + "litellm_provider": "watsonx", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 0.0005, - "output_cost_per_token": 0.002, - "litellm_provider": "watsonx", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 0.002, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/google/flan-t5-xl-3b": { - "max_tokens": 8192, + "input_cost_per_token": 6e-07, + "litellm_provider": "watsonx", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 6e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "watsonx", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 6e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/ibm/granite-13b-chat-v2": { - "max_tokens": 8192, + "input_cost_per_token": 6e-07, + "litellm_provider": "watsonx", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 6e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "watsonx", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 6e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/ibm/granite-13b-instruct-v2": { - "max_tokens": 8192, + "input_cost_per_token": 6e-07, + "litellm_provider": "watsonx", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 6e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "watsonx", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 6e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/ibm/granite-3-3-8b-instruct": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "watsonx", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "watsonx", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 2e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": false }, "watsonx/ibm/granite-4-h-small": { - "max_tokens": 20480, + "input_cost_per_token": 6e-08, + "litellm_provider": "watsonx", "max_input_tokens": 131072, "max_output_tokens": 20480, - "input_cost_per_token": 6e-08, - "output_cost_per_token": 2.5e-07, - "litellm_provider": "watsonx", + "max_tokens": 20480, "mode": "chat", + "output_cost_per_token": 2.5e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": false }, "watsonx/ibm/granite-guardian-3-2-2b": { - "max_tokens": 8192, + "input_cost_per_token": 1e-07, + "litellm_provider": "watsonx", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "watsonx", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/ibm/granite-guardian-3-3-8b": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "watsonx", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "watsonx", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 2e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/ibm/granite-ttm-1024-96-r2": { - "max_tokens": 512, + "input_cost_per_token": 3.8e-07, + "litellm_provider": "watsonx", "max_input_tokens": 512, "max_output_tokens": 512, - "input_cost_per_token": 3.8e-07, - "output_cost_per_token": 3.8e-07, - "litellm_provider": "watsonx", + "max_tokens": 512, "mode": "chat", + "output_cost_per_token": 3.8e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/ibm/granite-ttm-1536-96-r2": { - "max_tokens": 512, + "input_cost_per_token": 3.8e-07, + "litellm_provider": "watsonx", "max_input_tokens": 512, "max_output_tokens": 512, - "input_cost_per_token": 3.8e-07, - "output_cost_per_token": 3.8e-07, - "litellm_provider": "watsonx", + "max_tokens": 512, "mode": "chat", + "output_cost_per_token": 3.8e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/ibm/granite-ttm-512-96-r2": { - "max_tokens": 512, + "input_cost_per_token": 3.8e-07, + "litellm_provider": "watsonx", "max_input_tokens": 512, "max_output_tokens": 512, - "input_cost_per_token": 3.8e-07, - "output_cost_per_token": 3.8e-07, - "litellm_provider": "watsonx", + "max_tokens": 512, "mode": "chat", + "output_cost_per_token": 3.8e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/ibm/granite-vision-3-2-2b": { - "max_tokens": 8192, + "input_cost_per_token": 1e-07, + "litellm_provider": "watsonx", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "watsonx", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": true }, "watsonx/meta-llama/llama-3-2-11b-vision-instruct": { - "max_tokens": 128000, + "input_cost_per_token": 3.5e-07, + "litellm_provider": "watsonx", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 3.5e-07, - "output_cost_per_token": 3.5e-07, - "litellm_provider": "watsonx", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.5e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": true }, "watsonx/meta-llama/llama-3-2-1b-instruct": { - "max_tokens": 128000, + "input_cost_per_token": 1e-07, + "litellm_provider": "watsonx", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "watsonx", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": false }, "watsonx/meta-llama/llama-3-2-3b-instruct": { - "max_tokens": 128000, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "watsonx", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 1.5e-07, - "litellm_provider": "watsonx", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1.5e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": false }, "watsonx/meta-llama/llama-3-2-90b-vision-instruct": { - "max_tokens": 128000, + "input_cost_per_token": 2e-06, + "litellm_provider": "watsonx", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 2e-06, - "output_cost_per_token": 2e-06, - "litellm_provider": "watsonx", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 2e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": true }, "watsonx/meta-llama/llama-3-3-70b-instruct": { - "max_tokens": 128000, + "input_cost_per_token": 7.1e-07, + "litellm_provider": "watsonx", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 7.1e-07, - "output_cost_per_token": 7.1e-07, - "litellm_provider": "watsonx", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 7.1e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": false }, "watsonx/meta-llama/llama-4-maverick-17b": { - "max_tokens": 128000, + "input_cost_per_token": 3.5e-07, + "litellm_provider": "watsonx", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 3.5e-07, - "output_cost_per_token": 1.4e-06, - "litellm_provider": "watsonx", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1.4e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": false }, "watsonx/meta-llama/llama-guard-3-11b-vision": { - "max_tokens": 128000, + "input_cost_per_token": 3.5e-07, + "litellm_provider": "watsonx", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 3.5e-07, - "output_cost_per_token": 3.5e-07, - "litellm_provider": "watsonx", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.5e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": true }, "watsonx/mistralai/mistral-medium-2505": { - "max_tokens": 128000, + "input_cost_per_token": 3e-06, + "litellm_provider": "watsonx", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 3e-06, - "output_cost_per_token": 1e-05, - "litellm_provider": "watsonx", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1e-05, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": false }, "watsonx/mistralai/mistral-small-2503": { - "max_tokens": 32000, + "input_cost_per_token": 1e-07, + "litellm_provider": "watsonx", "max_input_tokens": 32000, "max_output_tokens": 32000, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 3e-07, - "litellm_provider": "watsonx", + "max_tokens": 32000, "mode": "chat", + "output_cost_per_token": 3e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": false }, "watsonx/mistralai/mistral-small-3-1-24b-instruct-2503": { - "max_tokens": 32000, + "input_cost_per_token": 1e-07, + "litellm_provider": "watsonx", "max_input_tokens": 32000, "max_output_tokens": 32000, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 3e-07, - "litellm_provider": "watsonx", + "max_tokens": 32000, "mode": "chat", + "output_cost_per_token": 3e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": false }, "watsonx/mistralai/pixtral-12b-2409": { - "max_tokens": 128000, + "input_cost_per_token": 3.5e-07, + "litellm_provider": "watsonx", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 3.5e-07, - "output_cost_per_token": 3.5e-07, - "litellm_provider": "watsonx", + "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.5e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": true }, "watsonx/openai/gpt-oss-120b": { - "max_tokens": 8192, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "watsonx", "max_input_tokens": 131072, "max_output_tokens": 8192, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "watsonx", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 6e-07, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/sdaia/allam-1-13b-instruct": { - "max_tokens": 8192, + "input_cost_per_token": 1.8e-06, + "litellm_provider": "watsonx", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 1.8e-06, - "output_cost_per_token": 1.8e-06, - "litellm_provider": "watsonx", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.8e-06, "supports_function_calling": false, "supports_parallel_function_calling": false, "supports_vision": false }, "watsonx/whisper-large-v3-turbo": { "input_cost_per_second": 0.0001, - "output_cost_per_second": 0.0001, "litellm_provider": "watsonx", "mode": "audio_transcription", + "output_cost_per_second": 0.0001, "supported_endpoints": [ "/v1/audio/transcriptions" ] @@ -43046,6 +43050,7 @@ }, "xai/grok-3": { "cache_read_input_token_cost": 7.5e-07, + "deprecation_date": "2026-05-15", "input_cost_per_token": 3e-06, "litellm_provider": "xai", "max_input_tokens": 131072, @@ -43058,8 +43063,7 @@ "supports_prompt_caching": true, "supports_response_schema": false, "supports_tool_choice": true, - "supports_web_search": true, - "deprecation_date": "2026-05-15" + "supports_web_search": true }, "xai/grok-3-beta": { "cache_read_input_token_cost": 7.5e-07, @@ -43111,6 +43115,7 @@ }, "xai/grok-3-latest": { "cache_read_input_token_cost": 7.5e-07, + "deprecation_date": "2026-05-15", "input_cost_per_token": 3e-06, "litellm_provider": "xai", "max_input_tokens": 131072, @@ -43123,8 +43128,7 @@ "supports_prompt_caching": true, "supports_response_schema": false, "supports_tool_choice": true, - "supports_web_search": true, - "deprecation_date": "2026-05-15" + "supports_web_search": true }, "xai/grok-3-mini": { "cache_read_input_token_cost": 7.5e-08, @@ -43231,6 +43235,7 @@ "supports_web_search": true }, "xai/grok-4": { + "deprecation_date": "2026-05-15", "input_cost_per_token": 3e-06, "litellm_provider": "xai", "max_input_tokens": 256000, @@ -43242,11 +43247,11 @@ "supports_function_calling": true, "supports_prompt_caching": true, "supports_tool_choice": true, - "supports_web_search": true, - "deprecation_date": "2026-05-15" + "supports_web_search": true }, "xai/grok-4-fast-reasoning": { "cache_read_input_token_cost": 5e-08, + "deprecation_date": "2026-05-15", "input_cost_per_token": 2e-07, "input_cost_per_token_above_128k_tokens": 4e-07, "litellm_provider": "xai", @@ -43260,11 +43265,11 @@ "supports_function_calling": true, "supports_prompt_caching": true, "supports_tool_choice": true, - "supports_web_search": true, - "deprecation_date": "2026-05-15" + "supports_web_search": true }, "xai/grok-4-fast-non-reasoning": { "cache_read_input_token_cost": 5e-08, + "deprecation_date": "2026-05-15", "input_cost_per_token": 2e-07, "input_cost_per_token_above_128k_tokens": 4e-07, "litellm_provider": "xai", @@ -43278,10 +43283,10 @@ "supports_function_calling": true, "supports_prompt_caching": true, "supports_tool_choice": true, - "supports_web_search": true, - "deprecation_date": "2026-05-15" + "supports_web_search": true }, "xai/grok-4-0709": { + "deprecation_date": "2026-05-15", "input_cost_per_token": 3e-06, "input_cost_per_token_above_128k_tokens": 6e-06, "litellm_provider": "xai", @@ -43295,10 +43300,10 @@ "supports_function_calling": true, "supports_prompt_caching": true, "supports_tool_choice": true, - "supports_web_search": true, - "deprecation_date": "2026-05-15" + "supports_web_search": true }, "xai/grok-4-latest": { + "deprecation_date": "2026-05-15", "input_cost_per_token": 3e-06, "input_cost_per_token_above_128k_tokens": 6e-06, "litellm_provider": "xai", @@ -43312,8 +43317,7 @@ "supports_function_calling": true, "supports_prompt_caching": true, "supports_tool_choice": true, - "supports_web_search": true, - "deprecation_date": "2026-05-15" + "supports_web_search": true }, "xai/grok-4-1-fast": { "cache_read_input_token_cost": 5e-08, @@ -43338,6 +43342,7 @@ }, "xai/grok-4-1-fast-reasoning": { "cache_read_input_token_cost": 5e-08, + "deprecation_date": "2026-05-15", "input_cost_per_token": 2e-07, "input_cost_per_token_above_128k_tokens": 4e-07, "litellm_provider": "xai", @@ -43355,11 +43360,11 @@ "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "deprecation_date": "2026-05-15" + "supports_web_search": true }, "xai/grok-4-1-fast-reasoning-latest": { "cache_read_input_token_cost": 5e-08, + "deprecation_date": "2026-05-15", "input_cost_per_token": 2e-07, "input_cost_per_token_above_128k_tokens": 4e-07, "litellm_provider": "xai", @@ -43377,11 +43382,11 @@ "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "deprecation_date": "2026-05-15" + "supports_web_search": true }, "xai/grok-4-1-fast-non-reasoning": { "cache_read_input_token_cost": 5e-08, + "deprecation_date": "2026-05-15", "input_cost_per_token": 2e-07, "input_cost_per_token_above_128k_tokens": 4e-07, "litellm_provider": "xai", @@ -43398,11 +43403,11 @@ "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "deprecation_date": "2026-05-15" + "supports_web_search": true }, "xai/grok-4-1-fast-non-reasoning-latest": { "cache_read_input_token_cost": 5e-08, + "deprecation_date": "2026-05-15", "input_cost_per_token": 2e-07, "input_cost_per_token_above_128k_tokens": 4e-07, "litellm_provider": "xai", @@ -43419,91 +43424,90 @@ "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "deprecation_date": "2026-05-15" + "supports_web_search": true }, "xai/grok-4.20-multi-agent-beta-0309": { "cache_read_input_token_cost": 2e-07, + "cache_read_input_token_cost_above_200k_tokens": 4e-07, "input_cost_per_token": 1.25e-06, + "input_cost_per_token_above_200k_tokens": 2.5e-06, "litellm_provider": "xai", "max_input_tokens": 1000000, "max_output_tokens": 1000000, "max_tokens": 1000000, "mode": "chat", "output_cost_per_token": 2.5e-06, + "output_cost_per_token_above_200k_tokens": 5e-06, "source": "https://docs.x.ai/docs/models", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, + "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "input_cost_per_token_above_200k_tokens": 2.5e-06, - "output_cost_per_token_above_200k_tokens": 5e-06, - "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "supports_response_schema": true + "supports_web_search": true }, "xai/grok-4.20-beta-0309-reasoning": { "cache_read_input_token_cost": 2e-07, + "cache_read_input_token_cost_above_200k_tokens": 4e-07, "input_cost_per_token": 1.25e-06, + "input_cost_per_token_above_200k_tokens": 2.5e-06, "litellm_provider": "xai", "max_input_tokens": 1000000, "max_output_tokens": 1000000, "max_tokens": 1000000, "mode": "chat", "output_cost_per_token": 2.5e-06, + "output_cost_per_token_above_200k_tokens": 5e-06, "source": "https://docs.x.ai/docs/models", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, + "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "input_cost_per_token_above_200k_tokens": 2.5e-06, - "output_cost_per_token_above_200k_tokens": 5e-06, - "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "supports_response_schema": true + "supports_web_search": true }, "xai/grok-4.20-0309-reasoning": { "cache_read_input_token_cost": 2e-07, + "cache_read_input_token_cost_above_200k_tokens": 4e-07, "input_cost_per_token": 1.25e-06, + "input_cost_per_token_above_200k_tokens": 2.5e-06, "litellm_provider": "xai", "max_input_tokens": 1000000, "max_output_tokens": 1000000, "max_tokens": 1000000, "mode": "chat", "output_cost_per_token": 2.5e-06, + "output_cost_per_token_above_200k_tokens": 5e-06, "source": "https://docs.x.ai/docs/models", "supports_function_calling": true, + "supports_prompt_caching": true, "supports_reasoning": true, + "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "input_cost_per_token_above_200k_tokens": 2.5e-06, - "output_cost_per_token_above_200k_tokens": 5e-06, - "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "supports_prompt_caching": true, - "supports_response_schema": true + "supports_web_search": true }, "xai/grok-4.20-beta-0309-non-reasoning": { "cache_read_input_token_cost": 2e-07, + "cache_read_input_token_cost_above_200k_tokens": 4e-07, "input_cost_per_token": 1.25e-06, + "input_cost_per_token_above_200k_tokens": 2.5e-06, "litellm_provider": "xai", "max_input_tokens": 1000000, "max_output_tokens": 1000000, "max_tokens": 1000000, "mode": "chat", "output_cost_per_token": 2.5e-06, + "output_cost_per_token_above_200k_tokens": 5e-06, "source": "https://docs.x.ai/docs/models", "supports_function_calling": true, "supports_prompt_caching": true, + "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "input_cost_per_token_above_200k_tokens": 2.5e-06, - "output_cost_per_token_above_200k_tokens": 5e-06, - "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "supports_response_schema": true + "supports_web_search": true }, "xai/grok-4.3": { "cache_read_input_token_cost": 2e-07, @@ -43625,66 +43629,66 @@ }, "xai/grok-code-fast": { "cache_read_input_token_cost": 2e-07, + "cache_read_input_token_cost_above_200k_tokens": 4e-07, + "deprecation_date": "2026-05-15", "input_cost_per_token": 1e-06, + "input_cost_per_token_above_200k_tokens": 2e-06, "litellm_provider": "xai", "max_input_tokens": 256000, "max_output_tokens": 256000, "max_tokens": 256000, "mode": "chat", "output_cost_per_token": 2e-06, + "output_cost_per_token_above_200k_tokens": 4e-06, "source": "https://docs.x.ai/docs/models", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "input_cost_per_token_above_200k_tokens": 2e-06, - "output_cost_per_token_above_200k_tokens": 4e-06, - "cache_read_input_token_cost_above_200k_tokens": 4e-07, "supports_response_schema": true, - "supports_vision": true, - "deprecation_date": "2026-05-15" + "supports_tool_choice": true, + "supports_vision": true }, "xai/grok-code-fast-1": { "cache_read_input_token_cost": 2e-07, + "cache_read_input_token_cost_above_200k_tokens": 4e-07, + "deprecation_date": "2026-05-15", "input_cost_per_token": 1e-06, + "input_cost_per_token_above_200k_tokens": 2e-06, "litellm_provider": "xai", "max_input_tokens": 256000, "max_output_tokens": 256000, "max_tokens": 256000, "mode": "chat", "output_cost_per_token": 2e-06, + "output_cost_per_token_above_200k_tokens": 4e-06, "source": "https://docs.x.ai/docs/models", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "input_cost_per_token_above_200k_tokens": 2e-06, - "output_cost_per_token_above_200k_tokens": 4e-06, - "cache_read_input_token_cost_above_200k_tokens": 4e-07, "supports_response_schema": true, - "supports_vision": true, - "deprecation_date": "2026-05-15" + "supports_tool_choice": true, + "supports_vision": true }, "xai/grok-code-fast-1-0825": { "cache_read_input_token_cost": 2e-07, + "cache_read_input_token_cost_above_200k_tokens": 4e-07, + "deprecation_date": "2026-05-15", "input_cost_per_token": 1e-06, + "input_cost_per_token_above_200k_tokens": 2e-06, "litellm_provider": "xai", "max_input_tokens": 256000, "max_output_tokens": 256000, "max_tokens": 256000, "mode": "chat", "output_cost_per_token": 2e-06, + "output_cost_per_token_above_200k_tokens": 4e-06, "source": "https://docs.x.ai/docs/models", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "input_cost_per_token_above_200k_tokens": 2e-06, - "output_cost_per_token_above_200k_tokens": 4e-06, - "cache_read_input_token_cost_above_200k_tokens": 4e-07, "supports_response_schema": true, - "supports_vision": true, - "deprecation_date": "2026-05-15" + "supports_tool_choice": true, + "supports_vision": true }, "xai/grok-vision-beta": { "input_cost_per_image": 5e-06, @@ -43708,26 +43712,26 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "zai.glm-5": { "input_cost_per_token": 1e-06, - "output_cost_per_token": 3.2e-06, "litellm_provider": "bedrock_converse", "max_input_tokens": 200000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, - "supports_reasoning": true, "supports_native_structured_output": true, + "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "zai.glm-4.7-flash": { "input_cost_per_token": 7e-08, @@ -43737,179 +43741,179 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 4e-07, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "zai/glm-5": { "cache_creation_input_token_cost": 0, "cache_read_input_token_cost": 2e-07, "input_cost_per_token": 1e-06, - "output_cost_per_token": 3.2e-06, "litellm_provider": "zai", "max_input_tokens": 200000, "max_output_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.2e-06, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "zai/glm-5.1": { "cache_creation_input_token_cost": 0, "cache_read_input_token_cost": 2.6e-07, "input_cost_per_token": 1.4e-06, - "output_cost_per_token": 4.4e-06, "litellm_provider": "zai", "max_input_tokens": 200000, "max_output_tokens": 128000, "mode": "chat", + "output_cost_per_token": 4.4e-06, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "zai/glm-5-code": { "cache_creation_input_token_cost": 0, "cache_read_input_token_cost": 3e-07, "input_cost_per_token": 1.2e-06, - "output_cost_per_token": 5e-06, "litellm_provider": "zai", "max_input_tokens": 200000, "max_output_tokens": 128000, "mode": "chat", + "output_cost_per_token": 5e-06, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "zai/glm-4.7": { "cache_creation_input_token_cost": 0, "cache_read_input_token_cost": 1.1e-07, "input_cost_per_token": 6e-07, - "output_cost_per_token": 2.2e-06, "litellm_provider": "zai", "max_input_tokens": 200000, "max_output_tokens": 128000, "mode": "chat", + "output_cost_per_token": 2.2e-06, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "zai/glm-4.7-flash": { "cache_creation_input_token_cost": 0, "cache_read_input_token_cost": 0, "input_cost_per_token": 0, - "output_cost_per_token": 0, "litellm_provider": "zai", "max_input_tokens": 200000, "max_output_tokens": 128000, "mode": "chat", + "output_cost_per_token": 0, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "zai/glm-4.6": { "cache_creation_input_token_cost": 0, "cache_read_input_token_cost": 1.1e-07, "input_cost_per_token": 6e-07, - "output_cost_per_token": 2.2e-06, "litellm_provider": "zai", "max_input_tokens": 200000, "max_output_tokens": 128000, "mode": "chat", + "output_cost_per_token": 2.2e-06, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "zai/glm-4.5": { "input_cost_per_token": 6e-07, - "output_cost_per_token": 2.2e-06, "litellm_provider": "zai", "max_input_tokens": 128000, "max_output_tokens": 32000, "mode": "chat", + "output_cost_per_token": 2.2e-06, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "zai/glm-4.5v": { "input_cost_per_token": 6e-07, - "output_cost_per_token": 1.8e-06, "litellm_provider": "zai", "max_input_tokens": 128000, "max_output_tokens": 32000, "mode": "chat", + "output_cost_per_token": 1.8e-06, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, "supports_tool_choice": true, - "supports_vision": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_vision": true }, "zai/glm-4.5-x": { "input_cost_per_token": 2.2e-06, - "output_cost_per_token": 8.9e-06, "litellm_provider": "zai", "max_input_tokens": 128000, "max_output_tokens": 32000, "mode": "chat", + "output_cost_per_token": 8.9e-06, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "zai/glm-4.5-air": { "input_cost_per_token": 2e-07, - "output_cost_per_token": 1.1e-06, "litellm_provider": "zai", "max_input_tokens": 128000, "max_output_tokens": 32000, "mode": "chat", + "output_cost_per_token": 1.1e-06, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "zai/glm-4.5-airx": { "input_cost_per_token": 1.1e-06, - "output_cost_per_token": 4.5e-06, "litellm_provider": "zai", "max_input_tokens": 128000, "max_output_tokens": 32000, "mode": "chat", + "output_cost_per_token": 4.5e-06, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "zai/glm-4-32b-0414-128k": { "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, "litellm_provider": "zai", "max_input_tokens": 128000, "max_output_tokens": 32000, "mode": "chat", + "output_cost_per_token": 1e-07, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "zai/glm-4.5-flash": { "input_cost_per_token": 0, - "output_cost_per_token": 0, "litellm_provider": "zai", "max_input_tokens": 128000, "max_output_tokens": 32000, "mode": "chat", + "output_cost_per_token": 0, + "source": "https://docs.z.ai/guides/overview/pricing", "supports_function_calling": true, - "supports_tool_choice": true, - "source": "https://docs.z.ai/guides/overview/pricing" + "supports_tool_choice": true }, "vertex_ai/search_api": { "input_cost_per_query": 0.0015, @@ -44000,6 +44004,9 @@ }, "runwayml/gen4_turbo": { "litellm_provider": "runwayml", + "metadata": { + "comment": "5 credits per second @ $0.01 per credit = $0.05 per second" + }, "mode": "video_generation", "output_cost_per_video_per_second": 0.05, "source": "https://docs.dev.runwayml.com/guides/pricing/", @@ -44009,13 +44016,13 @@ ], "supported_output_modalities": [ "video" - ], - "metadata": { - "comment": "5 credits per second @ $0.01 per credit = $0.05 per second" - } + ] }, "runwayml/gen4_aleph": { "litellm_provider": "runwayml", + "metadata": { + "comment": "15 credits per second @ $0.01 per credit = $0.15 per second" + }, "mode": "video_generation", "output_cost_per_video_per_second": 0.15, "source": "https://docs.dev.runwayml.com/guides/pricing/", @@ -44025,13 +44032,13 @@ ], "supported_output_modalities": [ "video" - ], - "metadata": { - "comment": "15 credits per second @ $0.01 per credit = $0.15 per second" - } + ] }, "runwayml/gen3a_turbo": { "litellm_provider": "runwayml", + "metadata": { + "comment": "5 credits per second @ $0.01 per credit = $0.05 per second" + }, "mode": "video_generation", "output_cost_per_video_per_second": 0.05, "source": "https://docs.dev.runwayml.com/guides/pricing/", @@ -44041,15 +44048,15 @@ ], "supported_output_modalities": [ "video" - ], - "metadata": { - "comment": "5 credits per second @ $0.01 per credit = $0.05 per second" - } + ] }, "runwayml/gen4_image": { - "litellm_provider": "runwayml", - "mode": "image_generation", "input_cost_per_image": 0.05, + "litellm_provider": "runwayml", + "metadata": { + "comment": "5 credits per 720p image or 8 credits per 1080p image @ $0.01 per credit. Using 5 credits ($0.05) as base cost" + }, + "mode": "image_generation", "output_cost_per_image": 0.05, "source": "https://docs.dev.runwayml.com/guides/pricing/", "supported_modalities": [ @@ -44058,15 +44065,15 @@ ], "supported_output_modalities": [ "image" - ], - "metadata": { - "comment": "5 credits per 720p image or 8 credits per 1080p image @ $0.01 per credit. Using 5 credits ($0.05) as base cost" - } + ] }, "runwayml/gen4_image_turbo": { - "litellm_provider": "runwayml", - "mode": "image_generation", "input_cost_per_image": 0.02, + "litellm_provider": "runwayml", + "metadata": { + "comment": "2 credits per image (any resolution) @ $0.01 per credit = $0.02 per image" + }, + "mode": "image_generation", "output_cost_per_image": 0.02, "source": "https://docs.dev.runwayml.com/guides/pricing/", "supported_modalities": [ @@ -44075,1904 +44082,1901 @@ ], "supported_output_modalities": [ "image" - ], - "metadata": { - "comment": "2 credits per image (any resolution) @ $0.01 per credit = $0.02 per image" - } + ] }, "runwayml/eleven_multilingual_v2": { - "litellm_provider": "runwayml", - "mode": "audio_speech", "input_cost_per_character": 3e-07, - "source": "https://docs.dev.runwayml.com/guides/pricing/", + "litellm_provider": "runwayml", "metadata": { "comment": "Estimated cost based on standard TTS pricing. RunwayML uses ElevenLabs models." - } + }, + "mode": "audio_speech", + "source": "https://docs.dev.runwayml.com/guides/pricing/" }, "fireworks_ai/accounts/fireworks/models/qwen3-coder-480b-a35b-instruct": { - "max_tokens": 262144, + "input_cost_per_token": 4.5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 4.5e-07, - "output_cost_per_token": 1.8e-06, - "litellm_provider": "fireworks_ai", + "max_tokens": 262144, "mode": "chat", + "output_cost_per_token": 1.8e-06, "supports_reasoning": true }, "fireworks_ai/accounts/fireworks/models/flux-kontext-pro": { - "max_tokens": 4096, + "input_cost_per_token": 4e-08, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 4e-08, - "output_cost_per_token": 4e-08, - "litellm_provider": "fireworks_ai", - "mode": "image_generation" + "max_tokens": 4096, + "mode": "image_generation", + "output_cost_per_token": 4e-08 }, "fireworks_ai/accounts/fireworks/models/SSD-1B": { - "max_tokens": 4096, + "input_cost_per_token": 1.3e-10, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1.3e-10, - "output_cost_per_token": 1.3e-10, - "litellm_provider": "fireworks_ai", - "mode": "image_generation" + "max_tokens": 4096, + "mode": "image_generation", + "output_cost_per_token": 1.3e-10 }, "fireworks_ai/accounts/fireworks/models/chronos-hermes-13b-v2": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-13b": { - "max_tokens": 16384, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-13b-instruct": { - "max_tokens": 16384, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-13b-python": { - "max_tokens": 16384, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-34b": { - "max_tokens": 16384, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-34b-instruct": { - "max_tokens": 16384, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-34b-python": { - "max_tokens": 16384, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-70b": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-70b-instruct": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-70b-python": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-7b": { - "max_tokens": 16384, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-7b-instruct": { - "max_tokens": 16384, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/code-llama-7b-python": { - "max_tokens": 16384, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/code-qwen-1p5-7b": { - "max_tokens": 65536, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 65536, "max_output_tokens": 65536, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/codegemma-2b": { - "max_tokens": 8192, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/codegemma-7b": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/cogito-671b-v2-p1": { - "max_tokens": 163840, + "input_cost_per_token": 1.2e-06, + "litellm_provider": "fireworks_ai", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 1.2e-06, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 163840, + "mode": "chat", + "output_cost_per_token": 1.2e-06 }, "fireworks_ai/accounts/fireworks/models/cogito-v1-preview-llama-3b": { - "max_tokens": 131072, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/cogito-v1-preview-llama-70b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/cogito-v1-preview-llama-8b": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/cogito-v1-preview-qwen-14b": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/cogito-v1-preview-qwen-32b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/flux-kontext-max": { - "max_tokens": 4096, + "input_cost_per_token": 8e-08, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 8e-08, - "output_cost_per_token": 8e-08, - "litellm_provider": "fireworks_ai", - "mode": "image_generation" + "max_tokens": 4096, + "mode": "image_generation", + "output_cost_per_token": 8e-08 }, "fireworks_ai/accounts/fireworks/models/dbrx-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 1.2e-06, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1.2e-06, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1.2e-06 }, "fireworks_ai/accounts/fireworks/models/deepseek-coder-1b-base": { - "max_tokens": 16384, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-coder-33b-instruct": { - "max_tokens": 16384, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-coder-7b-base": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-coder-7b-base-v1p5": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-coder-7b-instruct-v1p5": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-coder-v2-lite-base": { - "max_tokens": 163840, + "input_cost_per_token": 5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 163840, + "mode": "chat", + "output_cost_per_token": 5e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-coder-v2-lite-instruct": { - "max_tokens": 163840, + "input_cost_per_token": 5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 163840, + "mode": "chat", + "output_cost_per_token": 5e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-prover-v2": { - "max_tokens": 163840, + "input_cost_per_token": 1.2e-06, + "litellm_provider": "fireworks_ai", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 1.2e-06, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 163840, + "mode": "chat", + "output_cost_per_token": 1.2e-06 }, "fireworks_ai/accounts/fireworks/models/deepseek-r1-0528-distill-qwen3-8b": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-r1-distill-llama-70b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-r1-distill-llama-8b": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-r1-distill-qwen-14b": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-r1-distill-qwen-1p5b": { - "max_tokens": 131072, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-r1-distill-qwen-32b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-r1-distill-qwen-7b": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-v2-lite-chat": { - "max_tokens": 163840, + "input_cost_per_token": 5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 163840, + "mode": "chat", + "output_cost_per_token": 5e-07 }, "fireworks_ai/accounts/fireworks/models/deepseek-v2p5": { - "max_tokens": 32768, + "input_cost_per_token": 1.2e-06, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1.2e-06, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1.2e-06 }, "fireworks_ai/accounts/fireworks/models/devstral-small-2505": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/dobby-mini-unhinged-plus-llama-3-1-8b": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/dobby-unhinged-llama-3-3-70b-new": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/dolphin-2-9-2-qwen2-72b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/dolphin-2p6-mixtral-8x7b": { - "max_tokens": 32768, + "input_cost_per_token": 5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 5e-07 }, "fireworks_ai/accounts/fireworks/models/ernie-4p5-21b-a3b-pt": { - "max_tokens": 4096, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/ernie-4p5-300b-a47b-pt": { - "max_tokens": 4096, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/fare-20b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/firefunction-v1": { - "max_tokens": 32768, + "input_cost_per_token": 5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 5e-07 }, "fireworks_ai/accounts/fireworks/models/firellava-13b": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/firesearch-ocr-v6": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/fireworks-asr-large": { - "max_tokens": 4096, + "input_cost_per_token": 0.0, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 0.0, - "output_cost_per_token": 0.0, - "litellm_provider": "fireworks_ai", - "mode": "audio_transcription" + "max_tokens": 4096, + "mode": "audio_transcription", + "output_cost_per_token": 0.0 }, "fireworks_ai/accounts/fireworks/models/fireworks-asr-v2": { - "max_tokens": 4096, + "input_cost_per_token": 0.0, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 0.0, - "output_cost_per_token": 0.0, - "litellm_provider": "fireworks_ai", - "mode": "audio_transcription" + "max_tokens": 4096, + "mode": "audio_transcription", + "output_cost_per_token": 0.0 }, "fireworks_ai/accounts/fireworks/models/flux-1-dev": { - "max_tokens": 4096, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/flux-1-dev-controlnet-union": { - "max_tokens": 4096, + "input_cost_per_token": 1e-09, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1e-09, - "output_cost_per_token": 1e-09, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1e-09 }, "fireworks_ai/accounts/fireworks/models/flux-1-dev-fp8": { - "max_tokens": 4096, + "input_cost_per_token": 5e-10, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 5e-10, - "output_cost_per_token": 5e-10, - "litellm_provider": "fireworks_ai", - "mode": "image_generation" + "max_tokens": 4096, + "mode": "image_generation", + "output_cost_per_token": 5e-10 }, "fireworks_ai/accounts/fireworks/models/flux-1-schnell": { - "max_tokens": 4096, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/flux-1-schnell-fp8": { - "max_tokens": 4096, + "input_cost_per_token": 3.5e-10, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 3.5e-10, - "output_cost_per_token": 3.5e-10, - "litellm_provider": "fireworks_ai", - "mode": "image_generation" + "max_tokens": 4096, + "mode": "image_generation", + "output_cost_per_token": 3.5e-10 }, "fireworks_ai/accounts/fireworks/models/gemma-2b-it": { - "max_tokens": 8192, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/gemma-3-27b-it": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/gemma-7b": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/gemma-7b-it": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/gemma2-9b-it": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/glm-4p5v": { - "max_tokens": 131072, + "input_cost_per_token": 1.2e-06, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1.2e-06, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "fireworks_ai", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 1.2e-06, "supports_reasoning": true }, "fireworks_ai/accounts/fireworks/models/gpt-oss-safeguard-120b": { - "max_tokens": 131072, + "input_cost_per_token": 1.2e-06, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1.2e-06, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 1.2e-06 }, "fireworks_ai/accounts/fireworks/models/gpt-oss-safeguard-20b": { - "max_tokens": 131072, + "input_cost_per_token": 5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 5e-07 }, "fireworks_ai/accounts/fireworks/models/hermes-2-pro-mistral-7b": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/internvl3-38b": { - "max_tokens": 16384, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/internvl3-78b": { - "max_tokens": 16384, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/internvl3-8b": { - "max_tokens": 16384, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/japanese-stable-diffusion-xl": { - "max_tokens": 4096, + "input_cost_per_token": 1.3e-10, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1.3e-10, - "output_cost_per_token": 1.3e-10, - "litellm_provider": "fireworks_ai", - "mode": "image_generation" + "max_tokens": 4096, + "mode": "image_generation", + "output_cost_per_token": 1.3e-10 }, "fireworks_ai/accounts/fireworks/models/kat-coder": { - "max_tokens": 262144, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/kat-dev-32b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/kat-dev-72b-exp": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/llama-guard-2-8b": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/llama-guard-3-1b": { - "max_tokens": 131072, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/llama-guard-3-8b": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v2-13b": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v2-13b-chat": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v2-70b": { - "max_tokens": 4096, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v2-70b-chat": { - "max_tokens": 2048, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 2048, "max_output_tokens": 2048, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 2048, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v2-7b": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v2-7b-chat": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v3-70b-instruct": { - "max_tokens": 8192, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v3-70b-instruct-hf": { - "max_tokens": 8192, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v3-8b": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v3-8b-instruct-hf": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v3p1-405b-instruct-long": { - "max_tokens": 4096, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v3p1-70b-instruct": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v3p1-70b-instruct-1b": { - "max_tokens": 4096, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v3p1-nemotron-70b-instruct": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v3p2-1b": { - "max_tokens": 131072, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v3p2-3b": { - "max_tokens": 131072, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/llama-v3p3-70b-instruct": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/llamaguard-7b": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/llava-yi-34b": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/minimax-m1-80k": { - "max_tokens": 4096, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/minimax-m2": { - "max_tokens": 4096, + "input_cost_per_token": 3e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1.2e-06 }, "fireworks_ai/accounts/fireworks/models/ministral-3-14b-instruct-2512": { - "max_tokens": 256000, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 256000, "max_output_tokens": 256000, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 256000, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/ministral-3-3b-instruct-2512": { - "max_tokens": 256000, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 256000, "max_output_tokens": 256000, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 256000, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/ministral-3-8b-instruct-2512": { - "max_tokens": 256000, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 256000, "max_output_tokens": 256000, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 256000, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/mistral-7b": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/mistral-7b-instruct-4k": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/mistral-7b-instruct-v0p2": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/mistral-7b-instruct-v3": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/mistral-7b-v0p2": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/mistral-large-3-fp8": { - "max_tokens": 256000, + "input_cost_per_token": 1.2e-06, + "litellm_provider": "fireworks_ai", "max_input_tokens": 256000, "max_output_tokens": 256000, - "input_cost_per_token": 1.2e-06, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 256000, + "mode": "chat", + "output_cost_per_token": 1.2e-06 }, "fireworks_ai/accounts/fireworks/models/mistral-nemo-base-2407": { - "max_tokens": 128000, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/mistral-nemo-instruct-2407": { - "max_tokens": 128000, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/mistral-small-24b-instruct-2501": { - "max_tokens": 32768, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/mixtral-8x22b": { - "max_tokens": 65536, + "input_cost_per_token": 1.2e-06, + "litellm_provider": "fireworks_ai", "max_input_tokens": 65536, "max_output_tokens": 65536, - "input_cost_per_token": 1.2e-06, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 1.2e-06 }, "fireworks_ai/accounts/fireworks/models/mixtral-8x22b-instruct": { - "max_tokens": 65536, + "input_cost_per_token": 1.2e-06, + "litellm_provider": "fireworks_ai", "max_input_tokens": 65536, "max_output_tokens": 65536, - "input_cost_per_token": 1.2e-06, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 1.2e-06 }, "fireworks_ai/accounts/fireworks/models/mixtral-8x7b": { - "max_tokens": 32768, + "input_cost_per_token": 5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 5e-07 }, "fireworks_ai/accounts/fireworks/models/mixtral-8x7b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 5e-07 }, "fireworks_ai/accounts/fireworks/models/mixtral-8x7b-instruct-hf": { - "max_tokens": 32768, + "input_cost_per_token": 5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 5e-07 }, "fireworks_ai/accounts/fireworks/models/mythomax-l2-13b": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/nemotron-nano-v2-12b-vl": { - "max_tokens": 4096, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/nous-capybara-7b-v1p9": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/nous-hermes-2-mixtral-8x7b-dpo": { - "max_tokens": 32768, + "input_cost_per_token": 5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 5e-07 }, "fireworks_ai/accounts/fireworks/models/nous-hermes-2-yi-34b": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/nous-hermes-llama2-13b": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/nous-hermes-llama2-70b": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/nous-hermes-llama2-7b": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/nvidia-nemotron-nano-12b-v2": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/nvidia-nemotron-nano-9b-v2": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/openchat-3p5-0106-7b": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/openhermes-2-mistral-7b": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/openhermes-2p5-mistral-7b": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/openorca-7b": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/phi-2-3b": { - "max_tokens": 2048, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 2048, "max_output_tokens": 2048, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 2048, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/phi-3-mini-128k-instruct": { - "max_tokens": 131072, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/phi-3-vision-128k-instruct": { - "max_tokens": 32064, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32064, "max_output_tokens": 32064, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32064, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/phind-code-llama-34b-python-v1": { - "max_tokens": 16384, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/phind-code-llama-34b-v1": { - "max_tokens": 16384, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/phind-code-llama-34b-v2": { - "max_tokens": 16384, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/playground-v2-1024px-aesthetic": { - "max_tokens": 4096, + "input_cost_per_token": 1.3e-10, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1.3e-10, - "output_cost_per_token": 1.3e-10, - "litellm_provider": "fireworks_ai", - "mode": "image_generation" + "max_tokens": 4096, + "mode": "image_generation", + "output_cost_per_token": 1.3e-10 }, "fireworks_ai/accounts/fireworks/models/playground-v2-5-1024px-aesthetic": { - "max_tokens": 4096, + "input_cost_per_token": 1.3e-10, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1.3e-10, - "output_cost_per_token": 1.3e-10, - "litellm_provider": "fireworks_ai", - "mode": "image_generation" + "max_tokens": 4096, + "mode": "image_generation", + "output_cost_per_token": 1.3e-10 }, "fireworks_ai/accounts/fireworks/models/pythia-12b": { - "max_tokens": 2048, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 2048, "max_output_tokens": 2048, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 2048, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen-qwq-32b-preview": { - "max_tokens": 32768, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen-v2p5-14b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen-v2p5-7b": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen1p5-72b-chat": { - "max_tokens": 32768, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2-7b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2-vl-2b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2-vl-72b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2-vl-7b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-0p5b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-14b": { - "max_tokens": 131072, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-1p5b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-32b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-32b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-72b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-72b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-7b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-0p5b": { - "max_tokens": 32768, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-0p5b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-14b": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-14b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-1p5b": { - "max_tokens": 32768, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-1p5b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-32b": { - "max_tokens": 32768, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-32b-instruct-128k": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-32b-instruct-32k-rope": { - "max_tokens": 32768, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-32b-instruct-64k": { - "max_tokens": 65536, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 65536, "max_output_tokens": 65536, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-3b": { - "max_tokens": 32768, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-3b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-7b": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-coder-7b-instruct": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-math-72b-instruct": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-vl-32b-instruct": { - "max_tokens": 128000, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-vl-3b-instruct": { - "max_tokens": 128000, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-vl-72b-instruct": { - "max_tokens": 128000, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen2p5-vl-7b-instruct": { - "max_tokens": 128000, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-0p6b": { - "max_tokens": 40960, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 40960, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-14b": { - "max_tokens": 40960, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 40960, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-1p7b": { - "max_tokens": 131072, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-1p7b-fp8-draft": { - "max_tokens": 262144, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-1p7b-fp8-draft-131072": { - "max_tokens": 131072, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-1p7b-fp8-draft-40960": { - "max_tokens": 40960, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 40960, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-235b-a22b": { - "max_tokens": 131072, + "input_cost_per_token": 2.2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 2.2e-07, - "output_cost_per_token": 8.8e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 8.8e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-235b-a22b-instruct-2507": { - "max_tokens": 262144, + "input_cost_per_token": 2.2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 2.2e-07, - "output_cost_per_token": 8.8e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 8.8e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-235b-a22b-thinking-2507": { - "max_tokens": 262144, + "input_cost_per_token": 2.2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 2.2e-07, - "output_cost_per_token": 8.8e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 8.8e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-30b-a3b": { - "max_tokens": 131072, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 6e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-30b-a3b-instruct-2507": { - "max_tokens": 262144, + "input_cost_per_token": 5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 5e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-30b-a3b-thinking-2507": { - "max_tokens": 262144, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-32b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", + "max_tokens": 131072, "mode": "chat", + "output_cost_per_token": 9e-07, "supports_reasoning": true }, "fireworks_ai/accounts/fireworks/models/qwen3-4b": { - "max_tokens": 40960, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 40960, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-4b-instruct-2507": { - "max_tokens": 262144, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-8b": { - "max_tokens": 40960, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", + "max_tokens": 40960, "mode": "chat", + "output_cost_per_token": 2e-07, "supports_reasoning": true }, "fireworks_ai/accounts/fireworks/models/qwen3-coder-30b-a3b-instruct": { - "max_tokens": 262144, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 6e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-coder-480b-instruct-bf16": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-embedding-0p6b": { - "max_tokens": 32768, + "input_cost_per_token": 0.0, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 0.0, - "output_cost_per_token": 0.0, - "litellm_provider": "fireworks_ai", - "mode": "embedding" + "max_tokens": 32768, + "mode": "embedding", + "output_cost_per_token": 0.0 }, "fireworks_ai/accounts/fireworks/models/qwen3-embedding-4b": { - "max_tokens": 40960, + "input_cost_per_token": 0.0, + "litellm_provider": "fireworks_ai", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 0.0, - "output_cost_per_token": 0.0, - "litellm_provider": "fireworks_ai", - "mode": "embedding" + "max_tokens": 40960, + "mode": "embedding", + "output_cost_per_token": 0.0 }, "fireworks_ai/accounts/fireworks/models/": { - "max_tokens": 40960, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 0.0, - "litellm_provider": "fireworks_ai", - "mode": "embedding" + "max_tokens": 40960, + "mode": "embedding", + "output_cost_per_token": 0.0 }, "fireworks_ai/accounts/fireworks/models/qwen3-next-80b-a3b-instruct": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-next-80b-a3b-thinking": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-reranker-0p6b": { - "max_tokens": 40960, + "input_cost_per_token": 0.0, + "litellm_provider": "fireworks_ai", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 0.0, - "output_cost_per_token": 0.0, - "litellm_provider": "fireworks_ai", - "mode": "rerank" + "max_tokens": 40960, + "mode": "rerank", + "output_cost_per_token": 0.0 }, "fireworks_ai/accounts/fireworks/models/qwen3-reranker-4b": { - "max_tokens": 40960, + "input_cost_per_token": 0.0, + "litellm_provider": "fireworks_ai", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 0.0, - "output_cost_per_token": 0.0, - "litellm_provider": "fireworks_ai", - "mode": "rerank" + "max_tokens": 40960, + "mode": "rerank", + "output_cost_per_token": 0.0 }, "fireworks_ai/accounts/fireworks/models/qwen3-reranker-8b": { - "max_tokens": 40960, + "input_cost_per_token": 0.0, + "litellm_provider": "fireworks_ai", "max_input_tokens": 40960, "max_output_tokens": 40960, - "input_cost_per_token": 0.0, - "output_cost_per_token": 0.0, - "litellm_provider": "fireworks_ai", - "mode": "rerank" + "max_tokens": 40960, + "mode": "rerank", + "output_cost_per_token": 0.0 }, "fireworks_ai/accounts/fireworks/models/qwen3-vl-235b-a22b-instruct": { - "max_tokens": 262144, + "input_cost_per_token": 2.2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 2.2e-07, - "output_cost_per_token": 8.8e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 8.8e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-vl-235b-a22b-thinking": { - "max_tokens": 262144, + "input_cost_per_token": 2.2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 2.2e-07, - "output_cost_per_token": 8.8e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 8.8e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-vl-30b-a3b-instruct": { - "max_tokens": 262144, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 6e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-vl-30b-a3b-thinking": { - "max_tokens": 262144, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 6e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-vl-32b-instruct": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3-vl-8b-instruct": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/qwen3p7-plus": { "cache_read_input_token_cost": 8e-08, @@ -45991,148 +45995,148 @@ "supports_vision": true }, "fireworks_ai/accounts/fireworks/models/qwq-32b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/rolm-ocr": { - "max_tokens": 128000, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/snorkel-mistral-7b-pairrm-dpo": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/stable-diffusion-xl-1024-v1-0": { - "max_tokens": 4096, + "input_cost_per_token": 1.3e-10, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1.3e-10, - "output_cost_per_token": 1.3e-10, - "litellm_provider": "fireworks_ai", - "mode": "image_generation" + "max_tokens": 4096, + "mode": "image_generation", + "output_cost_per_token": 1.3e-10 }, "fireworks_ai/accounts/fireworks/models/stablecode-3b": { - "max_tokens": 4096, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/starcoder-16b": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/starcoder-7b": { - "max_tokens": 8192, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 8192, "max_output_tokens": 8192, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/starcoder2-15b": { - "max_tokens": 16384, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/starcoder2-3b": { - "max_tokens": 16384, + "input_cost_per_token": 1e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 1e-07 }, "fireworks_ai/accounts/fireworks/models/starcoder2-7b": { - "max_tokens": 16384, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 16384, "max_output_tokens": 16384, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/toppy-m-7b": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/yi-34b": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/yi-34b-200k-capybara": { - "max_tokens": 200000, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 200000, "max_output_tokens": 200000, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 200000, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/yi-34b-chat": { - "max_tokens": 4096, + "input_cost_per_token": 9e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 9e-07, - "output_cost_per_token": 9e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 9e-07 }, "fireworks_ai/accounts/fireworks/models/yi-6b": { - "max_tokens": 4096, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 4096, "max_output_tokens": 4096, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 4096, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/models/zephyr-7b-beta": { - "max_tokens": 32768, + "input_cost_per_token": 2e-07, + "litellm_provider": "fireworks_ai", "max_input_tokens": 32768, "max_output_tokens": 32768, - "input_cost_per_token": 2e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "fireworks_ai", - "mode": "chat" + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2e-07 }, "fireworks_ai/accounts/fireworks/routers/glm-5p1-fast": { "cache_read_input_token_cost": 5.2e-07, @@ -46203,8 +46207,8 @@ "mode": "chat", "output_cost_per_token": 1.5e-06, "supports_function_calling": true, - "supports_vision": true, - "supports_reasoning": true + "supports_reasoning": true, + "supports_vision": true }, "scaleway/qwen/qwen3-235b-a22b-instruct-2507": { "input_cost_per_token": 7.5e-07, @@ -46279,10 +46283,10 @@ "max_tokens": 16384, "mode": "chat", "output_cost_per_token": 7.5e-06, - "supports_reasoning": true, "supports_function_calling": true, - "supports_vision": true, - "supports_tool_choice": true + "supports_reasoning": true, + "supports_tool_choice": true, + "supports_vision": true }, "scaleway/mistralai/mistral-small-3.2-24b-instruct-2506": { "input_cost_per_token": 1.5e-07, @@ -46303,8 +46307,8 @@ "max_tokens": 4096, "mode": "chat", "output_cost_per_token": 2e-07, - "supports_vision": true, - "supports_function_calling": true + "supports_function_calling": true, + "supports_vision": true }, "scaleway/BAAI/bge-multilingual-gemma2": { "input_cost_per_token": 1e-07, @@ -46323,1491 +46327,1491 @@ "supports_function_calling": true }, "novita/deepseek/deepseek-v3.2": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 1.345e-07, "input_cost_per_token": 2.69e-07, - "output_cost_per_token": 4e-07, + "input_cost_per_token_cache_hit": 1.345e-07, + "litellm_provider": "novita", "max_input_tokens": 163840, "max_output_tokens": 65536, "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 4e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "cache_read_input_token_cost": 1.345e-07, - "input_cost_per_token_cache_hit": 1.345e-07, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/minimax/minimax-m2.1": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 3e-08, "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, + "input_cost_per_token_cache_hit": 3e-08, + "litellm_provider": "novita", "max_input_tokens": 204800, "max_output_tokens": 131072, "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 1.2e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, "supports_response_schema": true, - "cache_read_input_token_cost": 3e-08, - "input_cost_per_token_cache_hit": 3e-08 + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/zai-org/glm-4.7": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 1.1e-07, "input_cost_per_token": 6e-07, - "output_cost_per_token": 2.2e-06, + "input_cost_per_token_cache_hit": 1.1e-07, + "litellm_provider": "novita", "max_input_tokens": 204800, "max_output_tokens": 131072, "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2.2e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "cache_read_input_token_cost": 1.1e-07, - "input_cost_per_token_cache_hit": 1.1e-07, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/xiaomimimo/mimo-v2-flash": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 2e-08, "input_cost_per_token": 1.1e-07, - "output_cost_per_token": 3.3e-07, + "input_cost_per_token_cache_hit": 2e-08, + "litellm_provider": "novita", "max_input_tokens": 262144, "max_output_tokens": 32000, "max_tokens": 32000, + "mode": "chat", + "output_cost_per_token": 3.3e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "cache_read_input_token_cost": 2e-08, - "input_cost_per_token_cache_hit": 2e-08, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/zai-org/autoglm-phone-9b-multilingual": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 3.5e-08, - "output_cost_per_token": 1.38e-07, + "litellm_provider": "novita", "max_input_tokens": 65536, "max_output_tokens": 65536, "max_tokens": 65536, - "supports_vision": true, - "supports_system_messages": true + "mode": "chat", + "output_cost_per_token": 1.38e-07, + "supports_system_messages": true, + "supports_vision": true }, "novita/moonshotai/kimi-k2-thinking": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 6e-07, - "output_cost_per_token": 2.5e-06, + "litellm_provider": "novita", "max_input_tokens": 262144, "max_output_tokens": 262144, "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 2.5e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/minimax/minimax-m2": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 3e-08, "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, + "input_cost_per_token_cache_hit": 3e-08, + "litellm_provider": "novita", "max_input_tokens": 204800, "max_output_tokens": 131072, "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 1.2e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, - "cache_read_input_token_cost": 3e-08, - "input_cost_per_token_cache_hit": 3e-08, "supports_reasoning": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/paddlepaddle/paddleocr-vl": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2e-08, - "output_cost_per_token": 2e-08, + "litellm_provider": "novita", "max_input_tokens": 16384, "max_output_tokens": 16384, "max_tokens": 16384, - "supports_vision": true, - "supports_system_messages": true + "mode": "chat", + "output_cost_per_token": 2e-08, + "supports_system_messages": true, + "supports_vision": true }, "novita/deepseek/deepseek-v3.2-exp": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2.7e-07, - "output_cost_per_token": 4.1e-07, + "litellm_provider": "novita", "max_input_tokens": 163840, "max_output_tokens": 65536, "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 4.1e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, - "supports_response_schema": true, - "supports_reasoning": true - }, - "novita/qwen/qwen3-vl-235b-a22b-thinking": { - "litellm_provider": "novita", - "mode": "chat", - "input_cost_per_token": 9.8e-07, - "output_cost_per_token": 3.95e-06, - "max_input_tokens": 131072, - "max_output_tokens": 32768, - "max_tokens": 32768, - "supports_vision": true, - "supports_system_messages": true, "supports_reasoning": true, - "supports_function_calling": true, + "supports_response_schema": true, + "supports_system_messages": true, "supports_tool_choice": true }, - "novita/zai-org/glm-4.6v": { + "novita/qwen/qwen3-vl-235b-a22b-thinking": { + "input_cost_per_token": 9.8e-07, "litellm_provider": "novita", - "mode": "chat", - "input_cost_per_token": 3e-07, - "output_cost_per_token": 9e-07, "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 3.95e-06, + "supports_function_calling": true, + "supports_reasoning": true, + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "novita/zai-org/glm-4.6v": { + "cache_read_input_token_cost": 5.5e-08, + "input_cost_per_token": 3e-07, + "input_cost_per_token_cache_hit": 5.5e-08, + "litellm_provider": "novita", + "max_input_tokens": 131072, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 9e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_vision": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "cache_read_input_token_cost": 5.5e-08, - "input_cost_per_token_cache_hit": 5.5e-08, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true }, "novita/zai-org/glm-4.6": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 1.1e-07, "input_cost_per_token": 5.5e-07, - "output_cost_per_token": 2.2e-06, + "input_cost_per_token_cache_hit": 1.1e-07, + "litellm_provider": "novita", "max_input_tokens": 204800, "max_output_tokens": 131072, "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2.2e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "cache_read_input_token_cost": 1.1e-07, - "input_cost_per_token_cache_hit": 1.1e-07, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/kwaipilot/kat-coder-pro": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 6e-08, "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, + "input_cost_per_token_cache_hit": 6e-08, + "litellm_provider": "novita", "max_input_tokens": 256000, "max_output_tokens": 128000, "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1.2e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, "supports_response_schema": true, - "cache_read_input_token_cost": 6e-08, - "input_cost_per_token_cache_hit": 6e-08 + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/qwen/qwen3-next-80b-a3b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 1.5e-06, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1.5e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/qwen/qwen3-next-80b-a3b-thinking": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 1.5e-06, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1.5e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/deepseek/deepseek-ocr": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 3e-08, - "output_cost_per_token": 3e-08, + "litellm_provider": "novita", "max_input_tokens": 8192, "max_output_tokens": 8192, "max_tokens": 8192, - "supports_tool_choice": true, - "supports_vision": true, + "mode": "chat", + "output_cost_per_token": 3e-08, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true, + "supports_vision": true }, "novita/deepseek/deepseek-v3.1-terminus": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 1.35e-07, "input_cost_per_token": 2.7e-07, - "output_cost_per_token": 1e-06, + "input_cost_per_token_cache_hit": 1.35e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "cache_read_input_token_cost": 1.35e-07, - "input_cost_per_token_cache_hit": 1.35e-07, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/qwen/qwen3-vl-235b-a22b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.5e-06, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1.5e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true, + "supports_vision": true }, "novita/qwen/qwen3-max": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2.11e-06, - "output_cost_per_token": 8.45e-06, + "litellm_provider": "novita", "max_input_tokens": 262144, "max_output_tokens": 65536, "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 8.45e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/skywork/r1v4-lite": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2e-07, - "output_cost_per_token": 6e-07, + "litellm_provider": "novita", "max_input_tokens": 262144, "max_output_tokens": 65536, "max_tokens": 65536, - "supports_tool_choice": true, - "supports_vision": true, + "mode": "chat", + "output_cost_per_token": 6e-07, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true, + "supports_vision": true }, "novita/deepseek/deepseek-v3.1": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 1.35e-07, "input_cost_per_token": 2.7e-07, - "output_cost_per_token": 1e-06, + "input_cost_per_token_cache_hit": 1.35e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "cache_read_input_token_cost": 1.35e-07, - "input_cost_per_token_cache_hit": 1.35e-07, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/moonshotai/kimi-k2-0905": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 6e-07, - "output_cost_per_token": 2.5e-06, + "litellm_provider": "novita", "max_input_tokens": 262144, "max_output_tokens": 262144, "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 2.5e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/qwen/qwen3-coder-480b-a35b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 3.8e-07, - "output_cost_per_token": 1.55e-06, + "litellm_provider": "novita", "max_input_tokens": 262144, "max_output_tokens": 65536, "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 1.55e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/qwen/qwen3-coder-30b-a3b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 7e-08, - "output_cost_per_token": 2.7e-07, + "litellm_provider": "novita", "max_input_tokens": 160000, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2.7e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/openai/gpt-oss-120b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 5e-08, - "output_cost_per_token": 2.5e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2.5e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_vision": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true }, "novita/moonshotai/kimi-k2-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 5.7e-07, - "output_cost_per_token": 2.3e-06, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 131072, "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2.3e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/deepseek/deepseek-v3-0324": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 1.35e-07, "input_cost_per_token": 2.7e-07, - "output_cost_per_token": 1.12e-06, + "input_cost_per_token_cache_hit": 1.35e-07, + "litellm_provider": "novita", "max_input_tokens": 163840, "max_output_tokens": 163840, "max_tokens": 163840, + "mode": "chat", + "output_cost_per_token": 1.12e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, "supports_response_schema": true, - "cache_read_input_token_cost": 1.35e-07, - "input_cost_per_token_cache_hit": 1.35e-07 + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/zai-org/glm-4.5": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 1.1e-07, "input_cost_per_token": 6e-07, - "output_cost_per_token": 2.2e-06, + "input_cost_per_token_cache_hit": 1.1e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 98304, "max_tokens": 98304, + "mode": "chat", + "output_cost_per_token": 2.2e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_reasoning": true, "supports_system_messages": true, - "cache_read_input_token_cost": 1.1e-07, - "input_cost_per_token_cache_hit": 1.1e-07, - "supports_reasoning": true + "supports_tool_choice": true }, "novita/qwen/qwen3-235b-a22b-thinking-2507": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 3e-07, - "output_cost_per_token": 3e-06, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 3e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_reasoning": true, "supports_system_messages": true, - "supports_reasoning": true + "supports_tool_choice": true }, "novita/meta-llama/llama-3.1-8b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2e-08, - "output_cost_per_token": 5e-08, + "litellm_provider": "novita", "max_input_tokens": 16384, "max_output_tokens": 16384, "max_tokens": 16384, - "supports_system_messages": true, - "supports_response_schema": true + "mode": "chat", + "output_cost_per_token": 5e-08, + "supports_response_schema": true, + "supports_system_messages": true }, "novita/google/gemma-3-12b-it": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 5e-08, - "output_cost_per_token": 1e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 8192, "max_tokens": 8192, - "supports_tool_choice": true, - "supports_vision": true, + "mode": "chat", + "output_cost_per_token": 1e-07, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true, + "supports_vision": true }, "novita/zai-org/glm-4.5v": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 1.1e-07, "input_cost_per_token": 6e-07, - "output_cost_per_token": 1.8e-06, + "input_cost_per_token_cache_hit": 1.1e-07, + "litellm_provider": "novita", "max_input_tokens": 65536, "max_output_tokens": 16384, "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 1.8e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_vision": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "cache_read_input_token_cost": 1.1e-07, - "input_cost_per_token_cache_hit": 1.1e-07, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true }, "novita/openai/gpt-oss-20b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 4e-08, - "output_cost_per_token": 1.5e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, - "supports_tool_choice": true, - "supports_vision": true, - "supports_system_messages": true, + "mode": "chat", + "output_cost_per_token": 1.5e-07, + "supports_reasoning": true, "supports_response_schema": true, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true }, "novita/qwen/qwen3-235b-a22b-instruct-2507": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 9e-08, - "output_cost_per_token": 5.8e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 16384, "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 5.8e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/deepseek/deepseek-r1-distill-qwen-14b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 1.5e-07, + "litellm_provider": "novita", "max_input_tokens": 32768, "max_output_tokens": 16384, "max_tokens": 16384, - "supports_tool_choice": true, - "supports_system_messages": true, + "mode": "chat", + "output_cost_per_token": 1.5e-07, + "supports_reasoning": true, "supports_response_schema": true, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/meta-llama/llama-3.3-70b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1.35e-07, - "output_cost_per_token": 4e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 120000, "max_tokens": 120000, + "mode": "chat", + "output_cost_per_token": 4e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/qwen/qwen-2.5-72b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 3.8e-07, - "output_cost_per_token": 4e-07, + "litellm_provider": "novita", "max_input_tokens": 32000, "max_output_tokens": 8192, "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 4e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/mistralai/mistral-nemo": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 4e-08, - "output_cost_per_token": 1.7e-07, + "litellm_provider": "novita", "max_input_tokens": 60288, "max_output_tokens": 16000, "max_tokens": 16000, - "supports_tool_choice": true, + "mode": "chat", + "output_cost_per_token": 1.7e-07, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/minimaxai/minimax-m1-80k": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 5.5e-07, - "output_cost_per_token": 2.2e-06, + "litellm_provider": "novita", "max_input_tokens": 1000000, "max_output_tokens": 40000, "max_tokens": 40000, + "mode": "chat", + "output_cost_per_token": 2.2e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, "supports_reasoning": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/deepseek/deepseek-r1-0528": { - "litellm_provider": "novita", - "mode": "chat", + "cache_read_input_token_cost": 3.5e-07, "input_cost_per_token": 7e-07, - "output_cost_per_token": 2.5e-06, + "input_cost_per_token_cache_hit": 3.5e-07, + "litellm_provider": "novita", "max_input_tokens": 163840, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2.5e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "cache_read_input_token_cost": 3.5e-07, - "input_cost_per_token_cache_hit": 3.5e-07, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/deepseek/deepseek-r1-distill-qwen-32b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 3e-07, - "output_cost_per_token": 3e-07, + "litellm_provider": "novita", "max_input_tokens": 64000, "max_output_tokens": 32000, "max_tokens": 32000, - "supports_tool_choice": true, - "supports_system_messages": true, + "mode": "chat", + "output_cost_per_token": 3e-07, + "supports_reasoning": true, "supports_response_schema": true, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/meta-llama/llama-3-8b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 4e-08, - "output_cost_per_token": 4e-08, + "litellm_provider": "novita", "max_input_tokens": 8192, "max_output_tokens": 8192, "max_tokens": 8192, - "supports_system_messages": true, - "supports_response_schema": true + "mode": "chat", + "output_cost_per_token": 4e-08, + "supports_response_schema": true, + "supports_system_messages": true }, "novita/microsoft/wizardlm-2-8x22b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 6.2e-07, - "output_cost_per_token": 6.2e-07, + "litellm_provider": "novita", "max_input_tokens": 65535, "max_output_tokens": 8000, "max_tokens": 8000, - "supports_system_messages": true, - "supports_response_schema": true + "mode": "chat", + "output_cost_per_token": 6.2e-07, + "supports_response_schema": true, + "supports_system_messages": true }, "novita/deepseek/deepseek-r1-0528-qwen3-8b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 6e-08, - "output_cost_per_token": 9e-08, + "litellm_provider": "novita", "max_input_tokens": 128000, "max_output_tokens": 32000, "max_tokens": 32000, - "supports_system_messages": true, - "supports_reasoning": true + "mode": "chat", + "output_cost_per_token": 9e-08, + "supports_reasoning": true, + "supports_system_messages": true }, "novita/deepseek/deepseek-r1-distill-llama-70b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 8e-07, - "output_cost_per_token": 8e-07, + "litellm_provider": "novita", "max_input_tokens": 8192, "max_output_tokens": 8192, "max_tokens": 8192, - "supports_tool_choice": true, - "supports_system_messages": true, + "mode": "chat", + "output_cost_per_token": 8e-07, + "supports_reasoning": true, "supports_response_schema": true, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/meta-llama/llama-3-70b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 5.1e-07, - "output_cost_per_token": 7.4e-07, + "litellm_provider": "novita", "max_input_tokens": 8192, "max_output_tokens": 8000, "max_tokens": 8000, - "supports_tool_choice": true, + "mode": "chat", + "output_cost_per_token": 7.4e-07, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/qwen/qwen3-235b-a22b-fp8": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2e-07, - "output_cost_per_token": 8e-07, + "litellm_provider": "novita", "max_input_tokens": 40960, "max_output_tokens": 20000, "max_tokens": 20000, - "supports_system_messages": true, + "mode": "chat", + "output_cost_per_token": 8e-07, "supports_reasoning": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_system_messages": true }, "novita/meta-llama/llama-4-maverick-17b-128e-instruct-fp8": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2.7e-07, - "output_cost_per_token": 8.5e-07, + "litellm_provider": "novita", "max_input_tokens": 1048576, "max_output_tokens": 8192, "max_tokens": 8192, - "supports_vision": true, + "mode": "chat", + "output_cost_per_token": 8.5e-07, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_vision": true }, "novita/meta-llama/llama-4-scout-17b-16e-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1.8e-07, - "output_cost_per_token": 5.9e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 131072, "max_tokens": 131072, - "supports_vision": true, - "supports_system_messages": true + "mode": "chat", + "output_cost_per_token": 5.9e-07, + "supports_system_messages": true, + "supports_vision": true }, "novita/nousresearch/hermes-2-pro-llama-3-8b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1.4e-07, - "output_cost_per_token": 1.4e-07, + "litellm_provider": "novita", "max_input_tokens": 8192, "max_output_tokens": 8192, "max_tokens": 8192, - "supports_tool_choice": true, + "mode": "chat", + "output_cost_per_token": 1.4e-07, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/qwen/qwen2.5-vl-72b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 8e-07, - "output_cost_per_token": 8e-07, + "litellm_provider": "novita", "max_input_tokens": 32768, "max_output_tokens": 32768, "max_tokens": 32768, - "supports_vision": true, - "supports_system_messages": true + "mode": "chat", + "output_cost_per_token": 8e-07, + "supports_system_messages": true, + "supports_vision": true }, "novita/sao10k/l3-70b-euryale-v2.1": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1.48e-06, - "output_cost_per_token": 1.48e-06, + "litellm_provider": "novita", "max_input_tokens": 8192, "max_output_tokens": 8192, "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 1.48e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/baidu/ernie-4.5-21B-a3b-thinking": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 7e-08, - "output_cost_per_token": 2.8e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 65536, "max_tokens": 65536, - "supports_system_messages": true, - "supports_reasoning": true + "mode": "chat", + "output_cost_per_token": 2.8e-07, + "supports_reasoning": true, + "supports_system_messages": true }, "novita/sao10k/l3-8b-lunaris": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 5e-08, - "output_cost_per_token": 5e-08, + "litellm_provider": "novita", "max_input_tokens": 8192, "max_output_tokens": 8192, "max_tokens": 8192, - "supports_tool_choice": true, + "mode": "chat", + "output_cost_per_token": 5e-08, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/baichuan/baichuan-m2-32b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 7e-08, - "output_cost_per_token": 7e-08, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 131072, "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 7e-08, "supports_system_messages": true }, "novita/baidu/ernie-4.5-vl-424b-a47b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 4.2e-07, - "output_cost_per_token": 1.25e-06, + "litellm_provider": "novita", "max_input_tokens": 123000, "max_output_tokens": 16000, "max_tokens": 16000, - "supports_vision": true, + "mode": "chat", + "output_cost_per_token": 1.25e-06, + "supports_reasoning": true, "supports_system_messages": true, - "supports_reasoning": true + "supports_vision": true }, "novita/baidu/ernie-4.5-300b-a47b-paddle": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2.8e-07, - "output_cost_per_token": 1.1e-06, + "litellm_provider": "novita", "max_input_tokens": 123000, "max_output_tokens": 12000, "max_tokens": 12000, - "supports_tool_choice": true, + "mode": "chat", + "output_cost_per_token": 1.1e-06, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/deepseek/deepseek-prover-v2-671b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 7e-07, - "output_cost_per_token": 2.5e-06, + "litellm_provider": "novita", "max_input_tokens": 160000, "max_output_tokens": 160000, "max_tokens": 160000, + "mode": "chat", + "output_cost_per_token": 2.5e-06, "supports_system_messages": true }, "novita/qwen/qwen3-32b-fp8": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1e-07, - "output_cost_per_token": 4.5e-07, + "litellm_provider": "novita", "max_input_tokens": 40960, "max_output_tokens": 20000, "max_tokens": 20000, - "supports_system_messages": true, - "supports_reasoning": true + "mode": "chat", + "output_cost_per_token": 4.5e-07, + "supports_reasoning": true, + "supports_system_messages": true }, "novita/qwen/qwen3-30b-a3b-fp8": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 9e-08, - "output_cost_per_token": 4.5e-07, + "litellm_provider": "novita", "max_input_tokens": 40960, "max_output_tokens": 20000, "max_tokens": 20000, - "supports_system_messages": true, - "supports_reasoning": true, + "mode": "chat", + "output_cost_per_token": 4.5e-07, "supports_function_calling": true, + "supports_reasoning": true, + "supports_system_messages": true, "supports_tool_choice": true }, "novita/google/gemma-3-27b-it": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1.19e-07, - "output_cost_per_token": 2e-07, + "litellm_provider": "novita", "max_input_tokens": 98304, "max_output_tokens": 16384, "max_tokens": 16384, - "supports_vision": true, - "supports_system_messages": true + "mode": "chat", + "output_cost_per_token": 2e-07, + "supports_system_messages": true, + "supports_vision": true }, "novita/deepseek/deepseek-v3-turbo": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 4e-07, - "output_cost_per_token": 1.3e-06, + "litellm_provider": "novita", "max_input_tokens": 64000, "max_output_tokens": 16000, "max_tokens": 16000, + "mode": "chat", + "output_cost_per_token": 1.3e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/deepseek/deepseek-r1-turbo": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 7e-07, - "output_cost_per_token": 2.5e-06, + "litellm_provider": "novita", "max_input_tokens": 64000, "max_output_tokens": 16000, "max_tokens": 16000, + "mode": "chat", + "output_cost_per_token": 2.5e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, "supports_reasoning": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/Sao10K/L3-8B-Stheno-v3.2": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 5e-08, - "output_cost_per_token": 5e-08, + "litellm_provider": "novita", "max_input_tokens": 8192, "max_output_tokens": 32000, "max_tokens": 32000, + "mode": "chat", + "output_cost_per_token": 5e-08, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/gryphe/mythomax-l2-13b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 9e-08, - "output_cost_per_token": 9e-08, + "litellm_provider": "novita", "max_input_tokens": 4096, "max_output_tokens": 3200, "max_tokens": 3200, + "mode": "chat", + "output_cost_per_token": 9e-08, "supports_system_messages": true }, "novita/baidu/ernie-4.5-vl-28b-a3b-thinking": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 3.9e-07, - "output_cost_per_token": 3.9e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 65536, "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 3.9e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_vision": true, - "supports_system_messages": true, + "supports_reasoning": true, "supports_response_schema": true, - "supports_reasoning": true + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true }, "novita/qwen/qwen3-vl-8b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 8e-08, - "output_cost_per_token": 5e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 5e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true, + "supports_vision": true }, "novita/zai-org/glm-4.5-air": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1.3e-07, - "output_cost_per_token": 8.5e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 98304, "max_tokens": 98304, + "mode": "chat", + "output_cost_per_token": 8.5e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_reasoning": true, "supports_system_messages": true, - "supports_reasoning": true + "supports_tool_choice": true }, "novita/qwen/qwen3-vl-30b-a3b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2e-07, - "output_cost_per_token": 7e-07, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 7e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true, + "supports_vision": true }, "novita/qwen/qwen3-vl-30b-a3b-thinking": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2e-07, - "output_cost_per_token": 1e-06, + "litellm_provider": "novita", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 1e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_vision": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true, + "supports_vision": true }, "novita/qwen/qwen3-omni-30b-a3b-thinking": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2.5e-07, - "output_cost_per_token": 9.7e-07, + "litellm_provider": "novita", "max_input_tokens": 65536, "max_output_tokens": 16384, "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 9.7e-07, + "supports_audio_input": true, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_vision": true, - "supports_system_messages": true, - "supports_response_schema": true, "supports_reasoning": true, - "supports_audio_input": true + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true }, "novita/qwen/qwen3-omni-30b-a3b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2.5e-07, - "output_cost_per_token": 9.7e-07, + "litellm_provider": "novita", "max_input_tokens": 65536, "max_output_tokens": 16384, "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 9.7e-07, + "supports_audio_input": true, + "supports_audio_output": true, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_vision": true, - "supports_system_messages": true, "supports_response_schema": true, - "supports_audio_input": true, - "supports_audio_output": true + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true }, "novita/qwen/qwen-mt-plus": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 2.5e-07, - "output_cost_per_token": 7.5e-07, + "litellm_provider": "novita", "max_input_tokens": 16384, "max_output_tokens": 8192, "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 7.5e-07, "supports_system_messages": true }, "novita/baidu/ernie-4.5-vl-28b-a3b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1.4e-07, - "output_cost_per_token": 5.6e-07, + "litellm_provider": "novita", "max_input_tokens": 30000, "max_output_tokens": 8000, "max_tokens": 8000, + "mode": "chat", + "output_cost_per_token": 5.6e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_vision": true, + "supports_reasoning": true, "supports_system_messages": true, - "supports_reasoning": true + "supports_tool_choice": true, + "supports_vision": true }, "novita/baidu/ernie-4.5-21B-a3b": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 7e-08, - "output_cost_per_token": 2.8e-07, + "litellm_provider": "novita", "max_input_tokens": 120000, "max_output_tokens": 8000, "max_tokens": 8000, + "mode": "chat", + "output_cost_per_token": 2.8e-07, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/qwen/qwen3-8b-fp8": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 3.5e-08, - "output_cost_per_token": 1.38e-07, + "litellm_provider": "novita", "max_input_tokens": 128000, "max_output_tokens": 20000, "max_tokens": 20000, - "supports_system_messages": true, - "supports_reasoning": true + "mode": "chat", + "output_cost_per_token": 1.38e-07, + "supports_reasoning": true, + "supports_system_messages": true }, "novita/qwen/qwen3-4b-fp8": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 3e-08, - "output_cost_per_token": 3e-08, + "litellm_provider": "novita", "max_input_tokens": 128000, "max_output_tokens": 20000, "max_tokens": 20000, - "supports_system_messages": true, - "supports_reasoning": true, + "mode": "chat", + "output_cost_per_token": 3e-08, "supports_function_calling": true, + "supports_reasoning": true, + "supports_system_messages": true, "supports_tool_choice": true }, "novita/qwen/qwen2.5-7b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 7e-08, - "output_cost_per_token": 7e-08, + "litellm_provider": "novita", "max_input_tokens": 32000, "max_output_tokens": 32000, "max_tokens": 32000, + "mode": "chat", + "output_cost_per_token": 7e-08, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/meta-llama/llama-3.2-3b-instruct": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 3e-08, - "output_cost_per_token": 5e-08, + "litellm_provider": "novita", "max_input_tokens": 32768, "max_output_tokens": 32000, "max_tokens": 32000, + "mode": "chat", + "output_cost_per_token": 5e-08, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true + "supports_system_messages": true, + "supports_tool_choice": true }, "novita/sao10k/l31-70b-euryale-v2.2": { - "litellm_provider": "novita", - "mode": "chat", "input_cost_per_token": 1.48e-06, - "output_cost_per_token": 1.48e-06, + "litellm_provider": "novita", "max_input_tokens": 8192, "max_output_tokens": 8192, "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 1.48e-06, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_tool_choice": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_tool_choice": true }, "novita/qwen/qwen3-embedding-0.6b": { - "litellm_provider": "novita", - "mode": "embedding", "input_cost_per_token": 7e-08, - "output_cost_per_token": 0, + "litellm_provider": "novita", "max_input_tokens": 32768, "max_output_tokens": 32768, - "max_tokens": 32768 + "max_tokens": 32768, + "mode": "embedding", + "output_cost_per_token": 0 }, "novita/qwen/qwen3-embedding-8b": { - "litellm_provider": "novita", - "mode": "embedding", "input_cost_per_token": 7e-08, - "output_cost_per_token": 0, + "litellm_provider": "novita", "max_input_tokens": 32768, "max_output_tokens": 4096, - "max_tokens": 4096 + "max_tokens": 4096, + "mode": "embedding", + "output_cost_per_token": 0 }, "novita/baai/bge-m3": { - "litellm_provider": "novita", - "mode": "embedding", "input_cost_per_token": 1e-08, - "output_cost_per_token": 1e-08, + "litellm_provider": "novita", "max_input_tokens": 8192, "max_output_tokens": 96000, - "max_tokens": 96000 + "max_tokens": 96000, + "mode": "embedding", + "output_cost_per_token": 1e-08 }, "novita/qwen/qwen3-reranker-8b": { - "litellm_provider": "novita", - "mode": "rerank", "input_cost_per_token": 5e-08, - "output_cost_per_token": 5e-08, + "litellm_provider": "novita", "max_input_tokens": 32768, "max_output_tokens": 4096, - "max_tokens": 4096 + "max_tokens": 4096, + "mode": "rerank", + "output_cost_per_token": 5e-08 }, "novita/baai/bge-reranker-v2-m3": { - "litellm_provider": "novita", - "mode": "rerank", "input_cost_per_token": 1e-08, - "output_cost_per_token": 1e-08, + "litellm_provider": "novita", "max_input_tokens": 8000, "max_output_tokens": 8000, - "max_tokens": 8000 + "max_tokens": 8000, + "mode": "rerank", + "output_cost_per_token": 1e-08 }, "llamagate/llama-3.1-8b": { - "max_tokens": 8192, + "input_cost_per_token": 3e-08, + "litellm_provider": "llamagate", "max_input_tokens": 131072, "max_output_tokens": 8192, - "input_cost_per_token": 3e-08, - "output_cost_per_token": 5e-08, - "litellm_provider": "llamagate", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 5e-08, "supports_function_calling": true, "supports_response_schema": true }, "llamagate/llama-3.2-3b": { - "max_tokens": 8192, + "input_cost_per_token": 4e-08, + "litellm_provider": "llamagate", "max_input_tokens": 131072, "max_output_tokens": 8192, - "input_cost_per_token": 4e-08, - "output_cost_per_token": 8e-08, - "litellm_provider": "llamagate", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 8e-08, "supports_function_calling": true, "supports_response_schema": true }, "llamagate/mistral-7b-v0.3": { - "max_tokens": 8192, + "input_cost_per_token": 1e-07, + "litellm_provider": "llamagate", "max_input_tokens": 32768, "max_output_tokens": 8192, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 1.5e-07, - "litellm_provider": "llamagate", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.5e-07, "supports_function_calling": true, "supports_response_schema": true }, "llamagate/qwen3-8b": { - "max_tokens": 8192, + "input_cost_per_token": 4e-08, + "litellm_provider": "llamagate", "max_input_tokens": 32768, "max_output_tokens": 8192, - "input_cost_per_token": 4e-08, - "output_cost_per_token": 1.4e-07, - "litellm_provider": "llamagate", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.4e-07, "supports_function_calling": true, "supports_response_schema": true }, "llamagate/dolphin3-8b": { - "max_tokens": 8192, + "input_cost_per_token": 8e-08, + "litellm_provider": "llamagate", "max_input_tokens": 128000, "max_output_tokens": 8192, - "input_cost_per_token": 8e-08, - "output_cost_per_token": 1.5e-07, - "litellm_provider": "llamagate", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.5e-07, "supports_function_calling": true, "supports_response_schema": true }, "llamagate/deepseek-r1-8b": { - "max_tokens": 16384, + "input_cost_per_token": 1e-07, + "litellm_provider": "llamagate", "max_input_tokens": 65536, "max_output_tokens": 16384, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "llamagate", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 2e-07, "supports_function_calling": true, - "supports_response_schema": true, - "supports_reasoning": true + "supports_reasoning": true, + "supports_response_schema": true }, "llamagate/deepseek-r1-7b-qwen": { - "max_tokens": 16384, + "input_cost_per_token": 8e-08, + "litellm_provider": "llamagate", "max_input_tokens": 131072, "max_output_tokens": 16384, - "input_cost_per_token": 8e-08, - "output_cost_per_token": 1.5e-07, - "litellm_provider": "llamagate", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 1.5e-07, "supports_function_calling": true, - "supports_response_schema": true, - "supports_reasoning": true + "supports_reasoning": true, + "supports_response_schema": true }, "llamagate/openthinker-7b": { - "max_tokens": 8192, + "input_cost_per_token": 8e-08, + "litellm_provider": "llamagate", "max_input_tokens": 32768, "max_output_tokens": 8192, - "input_cost_per_token": 8e-08, - "output_cost_per_token": 1.5e-07, - "litellm_provider": "llamagate", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.5e-07, "supports_function_calling": true, - "supports_response_schema": true, - "supports_reasoning": true + "supports_reasoning": true, + "supports_response_schema": true }, "llamagate/qwen2.5-coder-7b": { - "max_tokens": 8192, + "input_cost_per_token": 6e-08, + "litellm_provider": "llamagate", "max_input_tokens": 32768, "max_output_tokens": 8192, - "input_cost_per_token": 6e-08, - "output_cost_per_token": 1.2e-07, - "litellm_provider": "llamagate", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 1.2e-07, "supports_function_calling": true, "supports_response_schema": true }, "llamagate/deepseek-coder-6.7b": { - "max_tokens": 4096, + "input_cost_per_token": 6e-08, + "litellm_provider": "llamagate", "max_input_tokens": 16384, "max_output_tokens": 4096, - "input_cost_per_token": 6e-08, - "output_cost_per_token": 1.2e-07, - "litellm_provider": "llamagate", + "max_tokens": 4096, "mode": "chat", + "output_cost_per_token": 1.2e-07, "supports_function_calling": true, "supports_response_schema": true }, "llamagate/codellama-7b": { - "max_tokens": 4096, + "input_cost_per_token": 6e-08, + "litellm_provider": "llamagate", "max_input_tokens": 16384, "max_output_tokens": 4096, - "input_cost_per_token": 6e-08, - "output_cost_per_token": 1.2e-07, - "litellm_provider": "llamagate", + "max_tokens": 4096, "mode": "chat", + "output_cost_per_token": 1.2e-07, "supports_function_calling": true, "supports_response_schema": true }, "llamagate/qwen3-vl-8b": { - "max_tokens": 8192, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "llamagate", "max_input_tokens": 32768, "max_output_tokens": 8192, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 5.5e-07, - "litellm_provider": "llamagate", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 5.5e-07, "supports_function_calling": true, "supports_response_schema": true, "supports_vision": true }, "llamagate/llava-7b": { - "max_tokens": 2048, + "input_cost_per_token": 1e-07, + "litellm_provider": "llamagate", "max_input_tokens": 4096, "max_output_tokens": 2048, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "llamagate", + "max_tokens": 2048, "mode": "chat", + "output_cost_per_token": 2e-07, "supports_response_schema": true, "supports_vision": true }, "llamagate/gemma3-4b": { - "max_tokens": 8192, + "input_cost_per_token": 3e-08, + "litellm_provider": "llamagate", "max_input_tokens": 128000, "max_output_tokens": 8192, - "input_cost_per_token": 3e-08, - "output_cost_per_token": 8e-08, - "litellm_provider": "llamagate", + "max_tokens": 8192, "mode": "chat", + "output_cost_per_token": 8e-08, "supports_function_calling": true, "supports_response_schema": true, "supports_vision": true }, "llamagate/nomic-embed-text": { - "max_tokens": 8192, - "max_input_tokens": 8192, "input_cost_per_token": 2e-08, - "output_cost_per_token": 0, "litellm_provider": "llamagate", - "mode": "embedding" + "max_input_tokens": 8192, + "max_tokens": 8192, + "mode": "embedding", + "output_cost_per_token": 0 }, "llamagate/qwen3-embedding-8b": { - "max_tokens": 40960, - "max_input_tokens": 40960, "input_cost_per_token": 2e-08, - "output_cost_per_token": 0, "litellm_provider": "llamagate", - "mode": "embedding" + "max_input_tokens": 40960, + "max_tokens": 40960, + "mode": "embedding", + "output_cost_per_token": 0 }, "libertai/hermes-3-8b-tee": { - "max_tokens": 16000, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "libertai", "max_input_tokens": 16000, "max_output_tokens": 16000, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "libertai", + "max_tokens": 16000, "mode": "chat", + "output_cost_per_token": 6e-07, + "source": "https://docs.libertai.io/apis/text/", "supports_function_calling": true, - "supports_tool_choice": true, "supports_system_messages": true, - "supports_vision": false, - "source": "https://docs.libertai.io/apis/text/" + "supports_tool_choice": true, + "supports_vision": false }, "libertai/gemma-4-31b-it": { - "max_tokens": 262144, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "libertai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "libertai", + "max_tokens": 262144, "mode": "chat", + "output_cost_per_token": 4e-07, + "source": "https://docs.libertai.io/apis/text/", "supports_function_calling": true, - "supports_tool_choice": true, "supports_system_messages": true, - "supports_vision": true, - "source": "https://docs.libertai.io/apis/text/" + "supports_tool_choice": true, + "supports_vision": true }, "libertai/gemma-4-31b-it-thinking": { - "max_tokens": 262144, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "libertai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 4e-07, - "litellm_provider": "libertai", + "max_tokens": 262144, "mode": "chat", + "output_cost_per_token": 4e-07, + "source": "https://docs.libertai.io/apis/text/", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, - "supports_vision": true, "supports_reasoning": true, - "source": "https://docs.libertai.io/apis/text/" + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true }, "libertai/qwen3.6-27b": { - "max_tokens": 262144, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "libertai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "libertai", + "max_tokens": 262144, "mode": "chat", + "output_cost_per_token": 5e-07, + "source": "https://docs.libertai.io/apis/text/", "supports_function_calling": true, - "supports_tool_choice": true, "supports_system_messages": true, - "supports_vision": true, - "source": "https://docs.libertai.io/apis/text/" + "supports_tool_choice": true, + "supports_vision": true }, "libertai/qwen3.6-27b-thinking": { - "max_tokens": 262144, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "libertai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "libertai", + "max_tokens": 262144, "mode": "chat", + "output_cost_per_token": 5e-07, + "source": "https://docs.libertai.io/apis/text/", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, - "supports_vision": true, "supports_reasoning": true, - "source": "https://docs.libertai.io/apis/text/" + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true }, "libertai/qwen3.6-35b-a3b": { - "max_tokens": 262144, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "libertai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "libertai", + "max_tokens": 262144, "mode": "chat", + "output_cost_per_token": 5e-07, + "source": "https://docs.libertai.io/apis/text/", "supports_function_calling": true, - "supports_tool_choice": true, "supports_system_messages": true, - "supports_vision": true, - "source": "https://docs.libertai.io/apis/text/" + "supports_tool_choice": true, + "supports_vision": true }, "libertai/qwen3.6-35b-a3b-thinking": { - "max_tokens": 262144, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "libertai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 5e-07, - "litellm_provider": "libertai", + "max_tokens": 262144, "mode": "chat", + "output_cost_per_token": 5e-07, + "source": "https://docs.libertai.io/apis/text/", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, - "supports_vision": true, "supports_reasoning": true, - "source": "https://docs.libertai.io/apis/text/" + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true }, "libertai/qwen3.5-122b-a10b": { - "max_tokens": 262144, + "input_cost_per_token": 2.5e-07, + "litellm_provider": "libertai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 2.5e-07, - "output_cost_per_token": 1.75e-06, - "litellm_provider": "libertai", + "max_tokens": 262144, "mode": "chat", + "output_cost_per_token": 1.75e-06, + "source": "https://docs.libertai.io/apis/text/", "supports_function_calling": true, - "supports_tool_choice": true, "supports_system_messages": true, - "supports_vision": true, - "source": "https://docs.libertai.io/apis/text/" + "supports_tool_choice": true, + "supports_vision": true }, "libertai/qwen3.5-122b-a10b-thinking": { - "max_tokens": 262144, + "input_cost_per_token": 2.5e-07, + "litellm_provider": "libertai", "max_input_tokens": 262144, "max_output_tokens": 262144, - "input_cost_per_token": 2.5e-07, - "output_cost_per_token": 1.75e-06, - "litellm_provider": "libertai", + "max_tokens": 262144, "mode": "chat", + "output_cost_per_token": 1.75e-06, + "source": "https://docs.libertai.io/apis/text/", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, - "supports_vision": true, "supports_reasoning": true, - "source": "https://docs.libertai.io/apis/text/" + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true }, "libertai/deepseek-v4-flash": { - "max_tokens": 200000, + "input_cost_per_token": 2.5e-07, + "litellm_provider": "libertai", "max_input_tokens": 200000, "max_output_tokens": 200000, - "input_cost_per_token": 2.5e-07, - "output_cost_per_token": 1.75e-06, - "litellm_provider": "libertai", + "max_tokens": 200000, "mode": "chat", + "output_cost_per_token": 1.75e-06, + "source": "https://docs.libertai.io/apis/text/", "supports_function_calling": true, - "supports_tool_choice": true, "supports_system_messages": true, - "supports_vision": false, - "source": "https://docs.libertai.io/apis/text/" + "supports_tool_choice": true, + "supports_vision": false }, "libertai/deepseek-v4-flash-thinking": { - "max_tokens": 200000, + "input_cost_per_token": 2.5e-07, + "litellm_provider": "libertai", "max_input_tokens": 200000, "max_output_tokens": 200000, - "input_cost_per_token": 2.5e-07, - "output_cost_per_token": 1.75e-06, - "litellm_provider": "libertai", + "max_tokens": 200000, "mode": "chat", + "output_cost_per_token": 1.75e-06, + "source": "https://docs.libertai.io/apis/text/", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_system_messages": true, - "supports_vision": false, "supports_reasoning": true, - "source": "https://docs.libertai.io/apis/text/" + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": false }, "libertai/bge-m3": { - "max_tokens": 8192, - "max_input_tokens": 8192, "input_cost_per_token": 1e-08, - "output_cost_per_token": 0.0, "litellm_provider": "libertai", + "max_input_tokens": 8192, + "max_tokens": 8192, "mode": "embedding", + "output_cost_per_token": 0.0, "source": "https://docs.libertai.io/apis/text/" }, "sarvam/sarvam-m": { @@ -47916,6 +47920,8 @@ "search_context_size_medium": 0.01 }, "supports_function_calling": true, + "supports_minimal_reasoning_effort": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -47924,9 +47930,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, - "supports_xhigh_reasoning_effort": false, - "supports_minimal_reasoning_effort": true + "supports_xhigh_reasoning_effort": false }, "gpt-5-search-api-2025-10-14": { "cache_read_input_token_cost": 1.25e-07, @@ -47943,6 +47947,7 @@ "search_context_size_medium": 0.01 }, "supports_function_calling": true, + "supports_none_reasoning_effort": false, "supports_parallel_function_calling": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -47951,7 +47956,6 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "supports_none_reasoning_effort": false, "supports_xhigh_reasoning_effort": false }, "gpt-realtime-mini-2025-10-06": { @@ -47992,6 +47996,7 @@ "cache_creation_input_audio_token_cost": 3e-07, "cache_read_input_audio_token_cost": 3e-07, "cache_read_input_token_cost": 6e-08, + "deprecation_date": "2027-01-20", "input_cost_per_audio_token": 1e-05, "input_cost_per_image": 8e-07, "input_cost_per_token": 6e-07, @@ -48019,8 +48024,7 @@ "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_system_messages": true, - "supports_tool_choice": true, - "deprecation_date": "2027-01-20" + "supports_tool_choice": true }, "gpt-realtime-whisper": { "input_cost_per_second": 0.0002833333333333333, @@ -48122,6 +48126,7 @@ "mode": "image_generation", "output_cost_per_image": 0.039, "output_cost_per_token": 0.0, + "rpm": 10, "source": "https://ai.google.dev/pricing", "supported_modalities": [ "text", @@ -48132,8 +48137,7 @@ "image" ], "supports_vision": true, - "tpm": 250000, - "rpm": 10 + "tpm": 250000 }, "gemini/gemini-2.0-flash-lite-001": { "cache_read_input_token_cost": 1.875e-08, @@ -48146,6 +48150,11 @@ "mode": "chat", "output_cost_per_token": 3e-07, "rpm": 4000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-2.0-flash-lite", "supported_modalities": [ "text", @@ -48164,14 +48173,10 @@ "supports_tool_choice": true, "supports_vision": true, "supports_web_search": true, - "tpm": 4000000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 4000000 }, "gemini-2.5-flash-native-audio-latest": { + "gemini_native_audio": true, "input_cost_per_audio_token": 1e-06, "input_cost_per_token": 3e-07, "litellm_provider": "gemini", @@ -48193,10 +48198,10 @@ "audio" ], "supports_audio_input": true, - "supports_audio_output": true, - "gemini_native_audio": true + "supports_audio_output": true }, "gemini-2.5-flash-native-audio-preview-09-2025": { + "gemini_native_audio": true, "input_cost_per_audio_token": 1e-06, "input_cost_per_token": 3e-07, "litellm_provider": "gemini", @@ -48218,10 +48223,10 @@ "audio" ], "supports_audio_input": true, - "supports_audio_output": true, - "gemini_native_audio": true + "supports_audio_output": true }, "gemini-2.5-flash-native-audio-preview-12-2025": { + "gemini_native_audio": true, "input_cost_per_audio_token": 3e-06, "input_cost_per_token": 5e-07, "litellm_provider": "gemini", @@ -48229,6 +48234,7 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "realtime", + "output_cost_per_audio_token": 1.2e-05, "output_cost_per_token": 2e-06, "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_endpoints": [ @@ -48243,11 +48249,10 @@ "audio" ], "supports_audio_input": true, - "supports_audio_output": true, - "gemini_native_audio": true, - "output_cost_per_audio_token": 1.2e-05 + "supports_audio_output": true }, "gemini-3.1-flash-live-preview": { + "gemini_audio_only_live": true, "input_cost_per_audio_token": 3e-06, "input_cost_per_image_token": 1e-06, "input_cost_per_token": 7.5e-07, @@ -48277,10 +48282,10 @@ "supports_audio_output": true, "supports_function_calling": true, "supports_vision": true, - "supports_web_search": true, - "gemini_audio_only_live": true + "supports_web_search": true }, "gemini/gemini-2.5-flash-native-audio-latest": { + "gemini_native_audio": true, "input_cost_per_audio_token": 1e-06, "input_cost_per_token": 3e-07, "litellm_provider": "gemini", @@ -48289,6 +48294,7 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 2.5e-06, + "rpm": 10, "source": "https://ai.google.dev/pricing", "supported_endpoints": [ "/v1/realtime" @@ -48303,11 +48309,10 @@ ], "supports_audio_input": true, "supports_audio_output": true, - "tpm": 250000, - "rpm": 10, - "gemini_native_audio": true + "tpm": 250000 }, "gemini/gemini-2.5-flash-native-audio-preview-09-2025": { + "gemini_native_audio": true, "input_cost_per_audio_token": 1e-06, "input_cost_per_token": 3e-07, "litellm_provider": "gemini", @@ -48316,6 +48321,7 @@ "max_tokens": 8192, "mode": "chat", "output_cost_per_token": 2.5e-06, + "rpm": 10, "source": "https://ai.google.dev/pricing", "supported_endpoints": [ "/v1/realtime" @@ -48330,11 +48336,10 @@ ], "supports_audio_input": true, "supports_audio_output": true, - "tpm": 250000, - "rpm": 10, - "gemini_native_audio": true + "tpm": 250000 }, "gemini/gemini-2.5-flash-native-audio-preview-12-2025": { + "gemini_native_audio": true, "input_cost_per_audio_token": 3e-06, "input_cost_per_token": 5e-07, "litellm_provider": "gemini", @@ -48342,7 +48347,9 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "realtime", + "output_cost_per_audio_token": 1.2e-05, "output_cost_per_token": 2e-06, + "rpm": 10, "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_endpoints": [ "/v1/realtime" @@ -48357,12 +48364,10 @@ ], "supports_audio_input": true, "supports_audio_output": true, - "tpm": 250000, - "rpm": 10, - "gemini_native_audio": true, - "output_cost_per_audio_token": 1.2e-05 + "tpm": 250000 }, "gemini/gemini-3.1-flash-live-preview": { + "gemini_audio_only_live": true, "input_cost_per_audio_token": 3e-06, "input_cost_per_image_token": 1e-06, "input_cost_per_token": 7.5e-07, @@ -48374,6 +48379,7 @@ "mode": "realtime", "output_cost_per_audio_token": 1.2e-05, "output_cost_per_token": 4.5e-06, + "rpm": 10, "source": "https://ai.google.dev/gemini-api/docs/pricing", "supported_endpoints": [ "/v1/realtime" @@ -48393,9 +48399,7 @@ "supports_function_calling": true, "supports_vision": true, "supports_web_search": true, - "tpm": 250000, - "rpm": 10, - "gemini_audio_only_live": true + "tpm": 250000 }, "gemini/gemini-3.1-flash-tts-preview": { "input_cost_per_token": 1e-06, @@ -48405,12 +48409,12 @@ "max_tokens": 16384, "mode": "audio_speech", "output_cost_per_token": 2e-05, + "rpm": 10, "source": "https://ai.google.dev/gemini-api/docs/models/gemini-3.1-flash-tts-preview", "supported_endpoints": [ "/v1/audio/speech" ], - "tpm": 4000000, - "rpm": 10 + "tpm": 4000000 }, "gemini-2.5-flash-preview-tts": { "input_cost_per_token": 5e-07, @@ -48434,6 +48438,11 @@ "output_cost_per_reasoning_token": 2.5e-06, "output_cost_per_token": 2.5e-06, "rpm": 100000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-preview", "supported_endpoints": [ "/v1/chat/completions", @@ -48461,12 +48470,7 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 8000000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 8000000 }, "gemini-flash-lite-latest": { "cache_read_input_token_cost": 1e-08, @@ -48480,6 +48484,11 @@ "output_cost_per_reasoning_token": 4e-07, "output_cost_per_token": 4e-07, "rpm": 15, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-lite", "supported_endpoints": [ "/v1/chat/completions", @@ -48507,12 +48516,7 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 250000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 250000 }, "gemini-pro-latest": { "cache_read_input_token_cost": 1.25e-07, @@ -48527,6 +48531,11 @@ "output_cost_per_token": 1e-05, "output_cost_per_token_above_200k_tokens": 1.5e-05, "rpm": 2000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -48552,12 +48561,7 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "tpm": 800000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 800000 }, "gemini/gemini-pro-latest": { "cache_read_input_token_cost": 1.25e-07, @@ -48572,6 +48576,11 @@ "output_cost_per_token": 1e-05, "output_cost_per_token_above_200k_tokens": 1.5e-05, "rpm": 2000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing", "supported_endpoints": [ "/v1/chat/completions", @@ -48597,12 +48606,7 @@ "supports_video_input": true, "supports_vision": true, "supports_web_search": true, - "tpm": 800000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 800000 }, "gemini-exp-1206": { "cache_read_input_token_cost": 3e-08, @@ -48616,6 +48620,11 @@ "output_cost_per_reasoning_token": 2.5e-06, "output_cost_per_token": 2.5e-06, "rpm": 100000, + "search_context_cost_per_query": { + "search_context_size_low": 0.035, + "search_context_size_medium": 0.035, + "search_context_size_high": 0.035 + }, "source": "https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-preview", "supported_endpoints": [ "/v1/chat/completions", @@ -48643,20 +48652,13 @@ "supports_url_context": true, "supports_vision": true, "supports_web_search": true, - "tpm": 8000000, - "search_context_cost_per_query": { - "search_context_size_low": 0.035, - "search_context_size_medium": 0.035, - "search_context_size_high": 0.035 - } + "tpm": 8000000 }, "vertex_ai/claude-sonnet-5@default": { - "deprecation_date": "2026-12-24", - "regional_endpoint_uplift_multiplier": 1.1, - "supports_mid_conversation_system": true, "cache_creation_input_token_cost": 2.5e-06, "cache_creation_input_token_cost_above_1hr": 4e-06, "cache_read_input_token_cost": 2e-07, + "deprecation_date": "2026-12-24", "input_cost_per_token": 2e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 1000000, @@ -48664,6 +48666,8 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1e-05, + "prompt_cache_min_tokens": 1024, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, @@ -48673,6 +48677,8 @@ "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -48680,13 +48686,9 @@ "supports_sampling_params": false, "supports_tool_choice": true, "supports_vision": true, - "supports_xhigh_reasoning_effort": true, - "supports_max_reasoning_effort": true, - "prompt_cache_min_tokens": 1024 + "supports_xhigh_reasoning_effort": true }, "vertex_ai/claude-sonnet-4-6@default": { - "regional_endpoint_uplift_multiplier": 1.1, - "supports_adaptive_thinking": true, "cache_creation_input_token_cost": 3.75e-06, "cache_creation_input_token_cost_above_1hr": 6e-06, "cache_read_input_token_cost": 3e-07, @@ -48697,40 +48699,43 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 1.5e-05, - "supports_assistant_prefill": true, - "supports_computer_use": true, - "supports_function_calling": true, - "supports_pdf_input": true, - "supports_prompt_caching": true, - "supports_reasoning": true, - "supports_response_schema": true, - "supports_max_reasoning_effort": true, - "supports_tool_choice": true, - "supports_vision": true, + "prompt_cache_min_tokens": 1024, + "regional_endpoint_uplift_multiplier": 1.1, "search_context_cost_per_query": { "search_context_size_high": 0.01, "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "supports_adaptive_thinking": true, + "supports_assistant_prefill": true, + "supports_computer_use": true, + "supports_function_calling": true, + "supports_legacy_thinking": true, + "supports_max_reasoning_effort": true, "supports_output_config": true, - "prompt_cache_min_tokens": 1024 + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true }, "duckduckgo/search": { - "litellm_provider": "duckduckgo", - "mode": "search", "input_cost_per_query": 0.0, + "litellm_provider": "duckduckgo", "metadata": { "notes": "DuckDuckGo Instant Answer API is free and does not require an API key." - } + }, + "mode": "search" }, "bedrock_mantle/openai.gpt-oss-120b": { "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, "litellm_provider": "bedrock_mantle", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, "mode": "chat", + "output_cost_per_token": 6e-07, "supported_endpoints": [ "/v1/chat/completions", "/v1/responses" @@ -48743,12 +48748,12 @@ }, "bedrock_mantle/openai.gpt-oss-20b": { "input_cost_per_token": 7.5e-08, - "output_cost_per_token": 3e-07, "litellm_provider": "bedrock_mantle", "max_input_tokens": 131072, "max_output_tokens": 32768, "max_tokens": 32768, "mode": "chat", + "output_cost_per_token": 3e-07, "supported_endpoints": [ "/v1/chat/completions", "/v1/responses" @@ -48761,12 +48766,12 @@ }, "bedrock_mantle/openai.gpt-oss-safeguard-120b": { "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, "litellm_provider": "bedrock_mantle", "max_input_tokens": 131072, "max_output_tokens": 65536, "max_tokens": 65536, "mode": "chat", + "output_cost_per_token": 6e-07, "supported_endpoints": [ "/v1/chat/completions" ], @@ -48777,12 +48782,12 @@ }, "bedrock_mantle/openai.gpt-oss-safeguard-20b": { "input_cost_per_token": 7.5e-08, - "output_cost_per_token": 3e-07, "litellm_provider": "bedrock_mantle", "max_input_tokens": 131072, "max_output_tokens": 65536, "max_tokens": 65536, "mode": "chat", + "output_cost_per_token": 3e-07, "supported_endpoints": [ "/v1/chat/completions" ], @@ -48792,20 +48797,19 @@ "supports_tool_choice": true }, "bedrock_mantle/openai.gpt-5.6-sol": { - "input_cost_per_token": 5.5e-06, - "input_cost_per_token_above_272k_tokens": 1.1e-05, "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_272k_tokens": 1.375e-05, "cache_read_input_token_cost": 5.5e-07, "cache_read_input_token_cost_above_272k_tokens": 1.1e-06, - "output_cost_per_token": 3.3e-05, - "output_cost_per_token_above_272k_tokens": 4.95e-05, + "input_cost_per_token": 5.5e-06, + "input_cost_per_token_above_272k_tokens": 1.1e-05, "litellm_provider": "bedrock_mantle", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "responses", - "use_openai_responses_path": true, + "output_cost_per_token": 3.3e-05, + "output_cost_per_token_above_272k_tokens": 4.95e-05, "supported_endpoints": [ "/v1/chat/completions", "/v1/responses" @@ -48818,6 +48822,9 @@ "text" ], "supports_function_calling": true, + "supports_minimal_reasoning_effort": false, + "supports_native_streaming": true, + "supports_none_reasoning_effort": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, @@ -48825,25 +48832,22 @@ "supports_vision": true, "supports_web_search": true, "supports_xhigh_reasoning_effort": true, - "supports_minimal_reasoning_effort": false, - "supports_none_reasoning_effort": true, - "supports_native_streaming": true + "use_openai_responses_path": true }, "bedrock_mantle/openai.gpt-5.6-terra": { - "input_cost_per_token": 2.2e-06, - "input_cost_per_token_above_272k_tokens": 4.4e-06, "cache_creation_input_token_cost": 2.75e-06, "cache_creation_input_token_cost_above_272k_tokens": 5.5e-06, "cache_read_input_token_cost": 2.2e-07, "cache_read_input_token_cost_above_272k_tokens": 4.4e-07, - "output_cost_per_token": 1.32e-05, - "output_cost_per_token_above_272k_tokens": 1.98e-05, + "input_cost_per_token": 2.2e-06, + "input_cost_per_token_above_272k_tokens": 4.4e-06, "litellm_provider": "bedrock_mantle", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "responses", - "use_openai_responses_path": true, + "output_cost_per_token": 1.32e-05, + "output_cost_per_token_above_272k_tokens": 1.98e-05, "supported_endpoints": [ "/v1/chat/completions", "/v1/responses" @@ -48856,29 +48860,30 @@ "text" ], "supports_function_calling": true, + "supports_native_streaming": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_streaming": true + "use_openai_responses_path": true }, "bedrock_mantle/openai.gpt-5.6-luna": { - "input_cost_per_token": 2.2e-07, - "input_cost_per_token_above_272k_tokens": 4.4e-07, "cache_creation_input_token_cost": 2.75e-07, "cache_creation_input_token_cost_above_272k_tokens": 5.5e-07, "cache_read_input_token_cost": 2.2e-08, "cache_read_input_token_cost_above_272k_tokens": 4.4e-08, - "output_cost_per_token": 1.32e-06, - "output_cost_per_token_above_272k_tokens": 1.98e-06, + "input_cost_per_token": 2.2e-07, + "input_cost_per_token_above_272k_tokens": 4.4e-07, "litellm_provider": "bedrock_mantle", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "responses", - "use_openai_responses_path": true, + "output_cost_per_token": 1.32e-06, + "output_cost_per_token_above_272k_tokens": 1.98e-06, + "source": "https://aws.amazon.com/bedrock/", "supported_endpoints": [ "/v1/chat/completions", "/v1/responses" @@ -48891,29 +48896,29 @@ "text" ], "supports_function_calling": true, + "supports_native_streaming": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "source": "https://aws.amazon.com/bedrock/", "supports_web_search": true, - "supports_native_streaming": true + "use_openai_responses_path": true }, "us.openai.gpt-5.6-sol": { - "input_cost_per_token": 5.5e-06, - "input_cost_per_token_above_272k_tokens": 1.1e-05, "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_272k_tokens": 1.375e-05, "cache_read_input_token_cost": 5.5e-07, "cache_read_input_token_cost_above_272k_tokens": 1.1e-06, - "output_cost_per_token": 3.3e-05, - "output_cost_per_token_above_272k_tokens": 4.95e-05, + "input_cost_per_token": 5.5e-06, + "input_cost_per_token_above_272k_tokens": 1.1e-05, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.3e-05, + "output_cost_per_token_above_272k_tokens": 4.95e-05, "supported_modalities": [ "text", "image" @@ -48926,19 +48931,19 @@ "supports_vision": true }, "global.openai.gpt-5.6-sol": { - "input_cost_per_token": 5e-06, - "input_cost_per_token_above_272k_tokens": 1e-05, "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_272k_tokens": 1.25e-05, "cache_read_input_token_cost": 5e-07, "cache_read_input_token_cost_above_272k_tokens": 1e-06, - "output_cost_per_token": 3e-05, - "output_cost_per_token_above_272k_tokens": 4.5e-05, + "input_cost_per_token": 5e-06, + "input_cost_per_token_above_272k_tokens": 1e-05, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3e-05, + "output_cost_per_token_above_272k_tokens": 4.5e-05, "supported_modalities": [ "text", "image" @@ -48951,19 +48956,19 @@ "supports_vision": true }, "us.openai.gpt-5.6-terra": { - "input_cost_per_token": 2.2e-06, - "input_cost_per_token_above_272k_tokens": 4.4e-06, "cache_creation_input_token_cost": 2.75e-06, "cache_creation_input_token_cost_above_272k_tokens": 5.5e-06, "cache_read_input_token_cost": 2.2e-07, "cache_read_input_token_cost_above_272k_tokens": 4.4e-07, - "output_cost_per_token": 1.32e-05, - "output_cost_per_token_above_272k_tokens": 1.98e-05, + "input_cost_per_token": 2.2e-06, + "input_cost_per_token_above_272k_tokens": 4.4e-06, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1.32e-05, + "output_cost_per_token_above_272k_tokens": 1.98e-05, "supported_modalities": [ "text", "image" @@ -48976,19 +48981,19 @@ "supports_vision": true }, "global.openai.gpt-5.6-terra": { - "input_cost_per_token": 2e-06, - "input_cost_per_token_above_272k_tokens": 4e-06, "cache_creation_input_token_cost": 2.5e-06, "cache_creation_input_token_cost_above_272k_tokens": 5e-06, "cache_read_input_token_cost": 2e-07, "cache_read_input_token_cost_above_272k_tokens": 4e-07, - "output_cost_per_token": 1.2e-05, - "output_cost_per_token_above_272k_tokens": 1.8e-05, + "input_cost_per_token": 2e-06, + "input_cost_per_token_above_272k_tokens": 4e-06, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1.2e-05, + "output_cost_per_token_above_272k_tokens": 1.8e-05, "supported_modalities": [ "text", "image" @@ -49001,19 +49006,19 @@ "supports_vision": true }, "us.openai.gpt-5.6-luna": { - "input_cost_per_token": 2.2e-07, - "input_cost_per_token_above_272k_tokens": 4.4e-07, "cache_creation_input_token_cost": 2.75e-07, "cache_creation_input_token_cost_above_272k_tokens": 5.5e-07, "cache_read_input_token_cost": 2.2e-08, "cache_read_input_token_cost_above_272k_tokens": 4.4e-08, - "output_cost_per_token": 1.32e-06, - "output_cost_per_token_above_272k_tokens": 1.98e-06, + "input_cost_per_token": 2.2e-07, + "input_cost_per_token_above_272k_tokens": 4.4e-07, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1.32e-06, + "output_cost_per_token_above_272k_tokens": 1.98e-06, "supported_modalities": [ "text", "image" @@ -49026,19 +49031,19 @@ "supports_vision": true }, "global.openai.gpt-5.6-luna": { - "input_cost_per_token": 2e-07, - "input_cost_per_token_above_272k_tokens": 4e-07, "cache_creation_input_token_cost": 2.5e-07, "cache_creation_input_token_cost_above_272k_tokens": 5e-07, "cache_read_input_token_cost": 2e-08, "cache_read_input_token_cost_above_272k_tokens": 4e-08, - "output_cost_per_token": 1.2e-06, - "output_cost_per_token_above_272k_tokens": 1.8e-06, + "input_cost_per_token": 2e-07, + "input_cost_per_token_above_272k_tokens": 4e-07, "litellm_provider": "bedrock_converse", "max_input_tokens": 1000000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 1.2e-06, + "output_cost_per_token_above_272k_tokens": 1.8e-06, "supported_modalities": [ "text", "image" @@ -49051,15 +49056,14 @@ "supports_vision": true }, "bedrock_mantle/openai.gpt-5.5": { - "input_cost_per_token": 5.5e-06, "cache_read_input_token_cost": 5.5e-07, - "output_cost_per_token": 3.3e-05, + "input_cost_per_token": 5.5e-06, "litellm_provider": "bedrock_mantle", "max_input_tokens": 272000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "responses", - "use_openai_responses_path": true, + "output_cost_per_token": 3.3e-05, "supported_endpoints": [ "/v1/responses" ], @@ -49075,18 +49079,18 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true + "supports_vision": true, + "use_openai_responses_path": true }, "bedrock_mantle/openai.gpt-5.4": { - "input_cost_per_token": 2.75e-06, "cache_read_input_token_cost": 2.75e-07, - "output_cost_per_token": 1.65e-05, + "input_cost_per_token": 2.75e-06, "litellm_provider": "bedrock_mantle", "max_input_tokens": 272000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "responses", - "use_openai_responses_path": true, + "output_cost_per_token": 1.65e-05, "supported_endpoints": [ "/v1/responses" ], @@ -49102,17 +49106,17 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true + "supports_vision": true, + "use_openai_responses_path": true }, "bedrock_mantle/google.gemma-4-31b": { "input_cost_per_token": 1.4e-07, - "output_cost_per_token": 4e-07, "litellm_provider": "bedrock_mantle", "max_input_tokens": 256000, "max_output_tokens": 256000, "max_tokens": 256000, "mode": "chat", - "use_openai_responses_path": true, + "output_cost_per_token": 4e-07, "supported_endpoints": [ "/v1/chat/completions", "/v1/responses" @@ -49121,17 +49125,17 @@ "supports_parallel_function_calling": false, "supports_reasoning": true, "supports_tool_choice": true, - "supports_vision": true + "supports_vision": true, + "use_openai_responses_path": true }, "bedrock_mantle/google.gemma-4-26b-a4b": { "input_cost_per_token": 1.3e-07, - "output_cost_per_token": 4e-07, "litellm_provider": "bedrock_mantle", "max_input_tokens": 256000, "max_output_tokens": 256000, "max_tokens": 256000, "mode": "chat", - "use_openai_responses_path": true, + "output_cost_per_token": 4e-07, "supported_endpoints": [ "/v1/chat/completions", "/v1/responses" @@ -49140,17 +49144,17 @@ "supports_parallel_function_calling": false, "supports_reasoning": true, "supports_tool_choice": true, - "supports_vision": true + "supports_vision": true, + "use_openai_responses_path": true }, "bedrock_mantle/google.gemma-4-e2b": { "input_cost_per_token": 4e-08, - "output_cost_per_token": 8e-08, "litellm_provider": "bedrock_mantle", "max_input_tokens": 128000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", - "use_openai_responses_path": true, + "output_cost_per_token": 8e-08, "supported_endpoints": [ "/v1/chat/completions", "/v1/responses" @@ -49159,28 +49163,23 @@ "supports_parallel_function_calling": false, "supports_reasoning": true, "supports_tool_choice": true, - "supports_vision": true + "supports_vision": true, + "use_openai_responses_path": true }, "bedrock_mantle/xai.grok-4.3": { - "use_openai_responses_path": true, - "input_cost_per_token": 1.25e-06, - "output_cost_per_token": 2.5e-06, "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 1.25e-06, "litellm_provider": "bedrock_mantle", "max_input_tokens": 1000000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 2.5e-06, + "source": "https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-xai-grok-4-3.html", "supported_endpoints": [ "/v1/chat/completions", "/v1/responses" ], - "supports_function_calling": true, - "supports_reasoning": true, - "supports_response_schema": true, - "supports_tool_choice": true, - "supports_vision": true, - "source": "https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-xai-grok-4-3.html", "supported_modalities": [ "text", "image" @@ -49188,18 +49187,23 @@ "supported_output_modalities": [ "text" ], - "supports_none_reasoning_effort": true + "supports_function_calling": true, + "supports_none_reasoning_effort": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true, + "use_openai_responses_path": true }, "bedrock_mantle/xai.grok-4.6": { - "use_openai_responses_path": true, - "input_cost_per_token": 2.2e-06, - "output_cost_per_token": 6.6e-06, "cache_read_input_token_cost": 5.5e-07, + "input_cost_per_token": 2.2e-06, "litellm_provider": "bedrock_mantle", "max_input_tokens": 500000, "max_output_tokens": 500000, "max_tokens": 500000, "mode": "chat", + "output_cost_per_token": 6.6e-06, "supported_endpoints": [ "/v1/chat/completions", "/v1/responses" @@ -49209,17 +49213,18 @@ "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true + "supports_vision": true, + "use_openai_responses_path": true }, "us.xai.grok-4.6": { - "input_cost_per_token": 2.2e-06, - "output_cost_per_token": 6.6e-06, "cache_read_input_token_cost": 5.5e-07, + "input_cost_per_token": 2.2e-06, "litellm_provider": "bedrock_converse", "max_input_tokens": 500000, "max_output_tokens": 500000, "max_tokens": 500000, "mode": "chat", + "output_cost_per_token": 6.6e-06, "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -49227,14 +49232,14 @@ "supports_vision": true }, "global.xai.grok-4.6": { - "input_cost_per_token": 2e-06, - "output_cost_per_token": 6e-06, "cache_read_input_token_cost": 5e-07, + "input_cost_per_token": 2e-06, "litellm_provider": "bedrock_converse", "max_input_tokens": 500000, "max_output_tokens": 500000, "max_tokens": 500000, "mode": "chat", + "output_cost_per_token": 6e-06, "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -49395,33 +49400,33 @@ }, "bedrock/us-east-1/zai.glm-5": { "input_cost_per_token": 1e-06, - "output_cost_per_token": 3.2e-06, "litellm_provider": "bedrock", "max_input_tokens": 200000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, + "supports_native_structured_output": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_native_structured_output": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/us-west-2/zai.glm-5": { "input_cost_per_token": 1e-06, - "output_cost_per_token": 3.2e-06, "litellm_provider": "bedrock", "max_input_tokens": 200000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", + "output_cost_per_token": 3.2e-06, + "source": "https://aws.amazon.com/bedrock/pricing/", "supports_function_calling": true, + "supports_native_structured_output": true, "supports_reasoning": true, "supports_system_messages": true, - "supports_native_structured_output": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "supports_tool_choice": true }, "bedrock/us-gov-east-1/anthropic.claude-haiku-4-5-20251001-v1:0": { "cache_creation_input_token_cost": 1.5e-06, @@ -49429,25 +49434,25 @@ "cache_read_input_token_cost": 1.2e-07, "input_cost_per_token": 1.2e-06, "litellm_provider": "bedrock", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 6e-06, + "prompt_cache_min_tokens": 4096, "source": "https://aws.amazon.com/about-aws/whats-new/2025/10/claude-4-5-haiku-anthropic-amazon-bedrock", "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, + "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_pdf_input": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "bedrock/us-gov-west-1/anthropic.claude-haiku-4-5-20251001-v1:0": { "cache_creation_input_token_cost": 1.5e-06, @@ -49455,211 +49460,212 @@ "cache_read_input_token_cost": 1.2e-07, "input_cost_per_token": 1.2e-06, "litellm_provider": "bedrock", - "supports_tool_search": true, "max_input_tokens": 200000, "max_output_tokens": 64000, "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 6e-06, + "prompt_cache_min_tokens": 4096, "source": "https://aws.amazon.com/about-aws/whats-new/2025/10/claude-4-5-haiku-anthropic-amazon-bedrock", "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, + "supports_pdf_input": true, "supports_prompt_caching": true, "supports_reasoning": true, "supports_response_schema": true, "supports_tool_choice": true, - "supports_vision": true, - "supports_native_structured_output": true, - "supports_pdf_input": true, - "supports_parallel_tool_use_config": true, - "prompt_cache_min_tokens": 4096 + "supports_tool_search": true, + "supports_vision": true }, "snowflake/claude-sonnet-4-5": { - "max_tokens": 16384, + "cache_read_input_token_cost": 3e-07, + "input_cost_per_token": 3e-06, + "litellm_provider": "snowflake", "max_input_tokens": 200000, "max_output_tokens": 16384, - "input_cost_per_token": 3e-06, - "output_cost_per_token": 1.5e-05, - "cache_read_input_token_cost": 3e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 1.5e-05, + "prompt_cache_min_tokens": 1024, "supports_function_calling": true, - "supports_vision": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_response_schema": true, - "prompt_cache_min_tokens": 1024 + "supports_system_messages": true, + "supports_vision": true }, "snowflake/claude-sonnet-4-6": { - "supports_adaptive_thinking": true, - "max_tokens": 16384, + "cache_read_input_token_cost": 3e-07, + "input_cost_per_token": 3e-06, + "litellm_provider": "snowflake", "max_input_tokens": 200000, "max_output_tokens": 16384, - "input_cost_per_token": 3e-06, - "output_cost_per_token": 1.5e-05, - "cache_read_input_token_cost": 3e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 1.5e-05, + "prompt_cache_min_tokens": 1024, + "supports_adaptive_thinking": true, "supports_function_calling": true, - "supports_vision": true, + "supports_legacy_thinking": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_response_schema": true, - "prompt_cache_min_tokens": 1024 + "supports_system_messages": true, + "supports_vision": true }, "snowflake/claude-4-sonnet": { - "max_tokens": 16384, + "cache_read_input_token_cost": 3e-07, + "input_cost_per_token": 3e-06, + "litellm_provider": "snowflake", "max_input_tokens": 200000, "max_output_tokens": 16384, - "input_cost_per_token": 3e-06, - "output_cost_per_token": 1.5e-05, - "cache_read_input_token_cost": 3e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 1.5e-05, "supports_function_calling": true, - "supports_vision": true, "supports_prompt_caching": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_vision": true }, "snowflake/claude-4-opus": { - "max_tokens": 16384, + "cache_read_input_token_cost": 5e-07, + "input_cost_per_token": 5e-06, + "litellm_provider": "snowflake", "max_input_tokens": 200000, "max_output_tokens": 16384, - "input_cost_per_token": 5e-06, - "output_cost_per_token": 2.5e-05, - "cache_read_input_token_cost": 5e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 2.5e-05, "supports_function_calling": true, - "supports_vision": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_system_messages": true, + "supports_vision": true }, "snowflake/claude-haiku-4-5": { - "max_tokens": 16384, + "cache_read_input_token_cost": 1e-07, + "input_cost_per_token": 1e-06, + "litellm_provider": "snowflake", "max_input_tokens": 200000, "max_output_tokens": 16384, - "input_cost_per_token": 1e-06, - "output_cost_per_token": 5e-06, - "cache_read_input_token_cost": 1e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 5e-06, + "prompt_cache_min_tokens": 4096, "supports_function_calling": true, - "supports_vision": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_response_schema": true, - "prompt_cache_min_tokens": 4096 + "supports_system_messages": true, + "supports_vision": true }, "snowflake/claude-3-7-sonnet": { - "max_tokens": 16384, + "cache_read_input_token_cost": 3e-07, + "input_cost_per_token": 3e-06, + "litellm_provider": "snowflake", "max_input_tokens": 200000, "max_output_tokens": 16384, - "input_cost_per_token": 3e-06, - "output_cost_per_token": 1.5e-05, - "cache_read_input_token_cost": 3e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 1.5e-05, "supports_function_calling": true, - "supports_vision": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_system_messages": true, + "supports_vision": true }, "snowflake/openai-gpt-4.1": { - "max_tokens": 16384, + "cache_read_input_token_cost": 5e-07, + "input_cost_per_token": 2e-06, + "litellm_provider": "snowflake", "max_input_tokens": 300000, "max_output_tokens": 16384, - "input_cost_per_token": 2e-06, - "output_cost_per_token": 8e-06, - "cache_read_input_token_cost": 5e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 8e-06, "supports_function_calling": true, - "supports_vision": true, "supports_prompt_caching": true, + "supports_response_schema": true, "supports_system_messages": true, - "supports_response_schema": true + "supports_vision": true }, "snowflake/openai-gpt-5": { - "max_tokens": 16384, + "cache_read_input_token_cost": 1.25e-07, + "input_cost_per_token": 1.25e-06, + "litellm_provider": "snowflake", "max_input_tokens": 300000, "max_output_tokens": 16384, - "input_cost_per_token": 1.25e-06, - "output_cost_per_token": 1e-05, - "cache_read_input_token_cost": 1.25e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 1e-05, "supports_function_calling": true, - "supports_vision": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_system_messages": true, + "supports_vision": true }, "snowflake/openai-gpt-5-mini": { - "max_tokens": 16384, + "input_cost_per_token": 3e-07, + "litellm_provider": "snowflake", "max_input_tokens": 1000000, "max_output_tokens": 16384, - "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 1.2e-06, "supports_function_calling": true, - "supports_system_messages": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_system_messages": true }, "snowflake/openai-gpt-5-nano": { - "max_tokens": 16384, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "snowflake", "max_input_tokens": 5000000, "max_output_tokens": 16384, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 6e-07, "supports_function_calling": true, - "supports_system_messages": true, - "supports_response_schema": true + "supports_response_schema": true, + "supports_system_messages": true }, "snowflake/llama4-maverick": { - "max_tokens": 16384, + "input_cost_per_token": 2.4e-07, + "litellm_provider": "snowflake", "max_input_tokens": 128000, "max_output_tokens": 16384, - "input_cost_per_token": 2.4e-07, - "output_cost_per_token": 9.7e-07, - "litellm_provider": "snowflake", + "max_tokens": 16384, "mode": "chat", + "output_cost_per_token": 9.7e-07, "supports_function_calling": true, "supports_system_messages": true }, "snowflake/snowflake-arctic-embed-l-v2.0": { - "max_tokens": 8192, - "max_input_tokens": 8192, "input_cost_per_token": 7e-08, - "output_cost_per_token": 0.0, "litellm_provider": "snowflake", - "mode": "embedding" + "max_input_tokens": 8192, + "max_tokens": 8192, + "mode": "embedding", + "output_cost_per_token": 0.0 }, "snowflake/snowflake-arctic-embed-m-v2.0": { - "max_tokens": 8192, - "max_input_tokens": 8192, "input_cost_per_token": 7e-08, - "output_cost_per_token": 0.0, "litellm_provider": "snowflake", - "mode": "embedding" + "max_input_tokens": 8192, + "max_tokens": 8192, + "mode": "embedding", + "output_cost_per_token": 0.0 }, "soniox/stt-async-v4": { + "input_cost_per_second": 0.0, "litellm_provider": "soniox", "max_output_tokens": 8000, "max_tokens": 8000, - "input_cost_per_second": 0.0, - "output_cost_per_second": 2.77778e-05, "mode": "audio_transcription", + "output_cost_per_second": 2.77778e-05, "source": "https://soniox.com/pricing", "supported_endpoints": [ "/v1/audio/transcriptions" @@ -49667,12 +49673,12 @@ "supports_audio_input": true }, "soniox/stt-async-v5": { + "input_cost_per_second": 0.0, "litellm_provider": "soniox", "max_output_tokens": 8000, "max_tokens": 8000, - "input_cost_per_second": 0.0, - "output_cost_per_second": 2.77778e-05, "mode": "audio_transcription", + "output_cost_per_second": 2.77778e-05, "source": "https://soniox.com/pricing", "supported_endpoints": [ "/v1/audio/transcriptions" @@ -49680,163 +49686,163 @@ "supports_audio_input": true }, "tensormesh/Qwen/Qwen3.5-397B-A17B-FP8": { - "litellm_provider": "tensormesh", - "mode": "chat", - "input_cost_per_token": 6e-07, - "output_cost_per_token": 3.6e-06, "cache_read_input_token_cost": 0, + "input_cost_per_token": 6e-07, + "litellm_provider": "tensormesh", "max_input_tokens": 262144, "max_output_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 3.6e-06, + "source": "https://serverless.tensormesh.ai/v1/models/openrouter", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "source": "https://serverless.tensormesh.ai/v1/models/openrouter" + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "tensormesh/Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8": { - "litellm_provider": "tensormesh", - "mode": "chat", - "input_cost_per_token": 4.5e-07, - "output_cost_per_token": 1.8e-06, "cache_read_input_token_cost": 0, + "input_cost_per_token": 4.5e-07, + "litellm_provider": "tensormesh", "max_input_tokens": 262144, "max_output_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 1.8e-06, + "source": "https://serverless.tensormesh.ai/v1/models/openrouter", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true, "supports_prompt_caching": true, + "supports_response_schema": true, "supports_system_messages": true, - "source": "https://serverless.tensormesh.ai/v1/models/openrouter" + "supports_tool_choice": true }, "tensormesh/Qwen/Qwen3.6-27B-FP8": { - "litellm_provider": "tensormesh", - "mode": "chat", - "input_cost_per_token": 3.2e-07, - "output_cost_per_token": 3.2e-06, "cache_read_input_token_cost": 0, + "input_cost_per_token": 3.2e-07, + "litellm_provider": "tensormesh", "max_input_tokens": 262144, "max_output_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 3.2e-06, + "source": "https://serverless.tensormesh.ai/v1/models/openrouter", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "source": "https://serverless.tensormesh.ai/v1/models/openrouter" + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "tensormesh/lukealonso/GLM-5.1-NVFP4-MTP": { - "litellm_provider": "tensormesh", - "mode": "chat", - "input_cost_per_token": 1.4e-06, - "output_cost_per_token": 4.4e-06, "cache_read_input_token_cost": 0, + "input_cost_per_token": 1.4e-06, + "litellm_provider": "tensormesh", "max_input_tokens": 202752, "max_output_tokens": 202752, + "mode": "chat", + "output_cost_per_token": 4.4e-06, + "source": "https://serverless.tensormesh.ai/v1/models/openrouter", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "source": "https://serverless.tensormesh.ai/v1/models/openrouter" + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "tensormesh/deepseek-ai/DeepSeek-V4-Flash": { - "litellm_provider": "tensormesh", - "mode": "chat", - "input_cost_per_token": 1.4e-07, - "output_cost_per_token": 2.8e-07, "cache_read_input_token_cost": 0, + "input_cost_per_token": 1.4e-07, + "litellm_provider": "tensormesh", "max_input_tokens": 32768, "max_output_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 2.8e-07, + "source": "https://serverless.tensormesh.ai/v1/models/openrouter", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "source": "https://serverless.tensormesh.ai/v1/models/openrouter" + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "tensormesh/moonshotai/Kimi-K2.6": { - "litellm_provider": "tensormesh", - "mode": "chat", - "input_cost_per_token": 9.6e-07, - "output_cost_per_token": 4e-06, "cache_read_input_token_cost": 0, + "input_cost_per_token": 9.6e-07, + "litellm_provider": "tensormesh", "max_input_tokens": 32768, "max_output_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 4e-06, + "source": "https://serverless.tensormesh.ai/v1/models/openrouter", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "source": "https://serverless.tensormesh.ai/v1/models/openrouter" + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "tensormesh/MiniMaxAI/MiniMax-M2.5": { - "litellm_provider": "tensormesh", - "mode": "chat", - "input_cost_per_token": 3e-07, - "output_cost_per_token": 1.2e-06, "cache_read_input_token_cost": 0, + "input_cost_per_token": 3e-07, + "litellm_provider": "tensormesh", "max_input_tokens": 196608, "max_output_tokens": 196608, + "mode": "chat", + "output_cost_per_token": 1.2e-06, + "source": "https://serverless.tensormesh.ai/v1/models/openrouter", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "source": "https://serverless.tensormesh.ai/v1/models/openrouter" + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "tensormesh/google/gemma-4-31B-it": { - "litellm_provider": "tensormesh", - "mode": "chat", - "input_cost_per_token": 1.4e-07, - "output_cost_per_token": 5.6e-07, "cache_read_input_token_cost": 0, + "input_cost_per_token": 1.4e-07, + "litellm_provider": "tensormesh", "max_input_tokens": 32768, "max_output_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 5.6e-07, + "source": "https://serverless.tensormesh.ai/v1/models/openrouter", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "source": "https://serverless.tensormesh.ai/v1/models/openrouter" + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "tensormesh/openai/gpt-oss-120b": { - "litellm_provider": "tensormesh", - "mode": "chat", - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, "cache_read_input_token_cost": 0, + "input_cost_per_token": 1.5e-07, + "litellm_provider": "tensormesh", "max_input_tokens": 131072, "max_output_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 6e-07, + "source": "https://serverless.tensormesh.ai/v1/models/openrouter", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "source": "https://serverless.tensormesh.ai/v1/models/openrouter" + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "tensormesh/openai/gpt-oss-20b": { - "litellm_provider": "tensormesh", - "mode": "chat", - "input_cost_per_token": 7e-08, - "output_cost_per_token": 2.8e-07, "cache_read_input_token_cost": 0, + "input_cost_per_token": 7e-08, + "litellm_provider": "tensormesh", "max_input_tokens": 131072, "max_output_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 2.8e-07, + "source": "https://serverless.tensormesh.ai/v1/models/openrouter", "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true, "supports_prompt_caching": true, - "supports_system_messages": true, "supports_reasoning": true, - "source": "https://serverless.tensormesh.ai/v1/models/openrouter" + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true }, "deepseek-v4-flash": { "cache_creation_input_token_cost": 0.0, @@ -49962,10 +49968,10 @@ "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_prompt_caching": true, + "supports_reasoning": true, "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_reasoning": true, "supports_vision": false }, "tencent/deepseek-v4-flash": { @@ -49988,119 +49994,119 @@ "supports_native_streaming": true, "supports_parallel_function_calling": true, "supports_prompt_caching": true, + "supports_reasoning": true, "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_reasoning": true, "supports_vision": false }, "cognition/swe-1.6": { - "input_cost_per_token": 5e-07, - "output_cost_per_token": 2.5e-06, "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 5e-07, "litellm_provider": "cognition", "mode": "chat", + "output_cost_per_token": 2.5e-06, + "source": "https://docs.devin.ai/windsurf/plugins/cascade/models", "supports_function_calling": true, - "supports_prompt_caching": true, - "source": "https://docs.devin.ai/windsurf/plugins/cascade/models" + "supports_prompt_caching": true }, "cognition/swe-1.7": { - "input_cost_per_token": 5e-07, - "output_cost_per_token": 2.5e-06, "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 5e-07, "litellm_provider": "cognition", "mode": "chat", + "output_cost_per_token": 2.5e-06, + "source": "https://docs.devin.ai/desktop/models", "supports_function_calling": true, - "supports_prompt_caching": true, - "source": "https://docs.devin.ai/desktop/models" + "supports_prompt_caching": true }, "cognition/swe-1.7-lightning": { - "input_cost_per_token": 2.5e-06, - "output_cost_per_token": 1.25e-05, "cache_read_input_token_cost": 1e-06, + "input_cost_per_token": 2.5e-06, "litellm_provider": "cognition", "mode": "chat", + "output_cost_per_token": 1.25e-05, + "source": "https://docs.devin.ai/desktop/models", "supports_function_calling": true, - "supports_prompt_caching": true, - "source": "https://docs.devin.ai/desktop/models" + "supports_prompt_caching": true }, "pinstripes/ps/glm-4.5-air": { - "max_tokens": 128000, + "input_cost_per_token": 1.25e-07, + "litellm_provider": "pinstripes", "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 1.25e-07, - "output_cost_per_token": 4.5e-07, - "litellm_provider": "pinstripes", + "max_tokens": 128000, "mode": "chat", - "supports_function_calling": true, + "output_cost_per_token": 4.5e-07, + "source": "https://pinstripes.io/pricing", "supports_assistant_prefill": true, - "supports_reasoning": true, - "source": "https://pinstripes.io/pricing" + "supports_function_calling": true, + "supports_reasoning": true }, "pinstripes/ps/qwen3.6-35b-a3b": { - "max_tokens": 131072, + "input_cost_per_token": 1.4e-07, + "litellm_provider": "pinstripes", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 1.4e-07, - "output_cost_per_token": 4.5e-07, - "litellm_provider": "pinstripes", + "max_tokens": 131072, "mode": "chat", - "supports_function_calling": true, + "output_cost_per_token": 4.5e-07, + "source": "https://pinstripes.io/pricing", "supports_assistant_prefill": true, - "supports_reasoning": true, - "source": "https://pinstripes.io/pricing" + "supports_function_calling": true, + "supports_reasoning": true }, "pinstripes/ps/qwen3-30b-a3b": { - "max_tokens": 131072, + "input_cost_per_token": 9e-08, + "litellm_provider": "pinstripes", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 9e-08, - "output_cost_per_token": 2e-07, - "litellm_provider": "pinstripes", + "max_tokens": 131072, "mode": "chat", - "supports_function_calling": true, + "output_cost_per_token": 2e-07, + "source": "https://pinstripes.io/pricing", "supports_assistant_prefill": true, - "supports_reasoning": true, - "source": "https://pinstripes.io/pricing" + "supports_function_calling": true, + "supports_reasoning": true }, "pinstripes/ps/qwen3-coder-30b-a3b": { - "max_tokens": 131072, + "input_cost_per_token": 3e-07, + "litellm_provider": "pinstripes", "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 3e-07, - "output_cost_per_token": 6e-07, - "litellm_provider": "pinstripes", + "max_tokens": 131072, "mode": "chat", - "supports_function_calling": true, + "output_cost_per_token": 6e-07, + "source": "https://pinstripes.io/pricing", "supports_assistant_prefill": true, - "supports_reasoning": false, - "source": "https://pinstripes.io/pricing" + "supports_function_calling": true, + "supports_reasoning": false }, "pinstripes/ps/deepseek-v4-flash": { - "max_tokens": 163840, + "input_cost_per_token": 1e-07, + "litellm_provider": "pinstripes", "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 2e-07, - "litellm_provider": "pinstripes", + "max_tokens": 163840, "mode": "chat", - "supports_function_calling": true, + "output_cost_per_token": 2e-07, + "source": "https://pinstripes.io/pricing", "supports_assistant_prefill": true, - "supports_reasoning": true, - "source": "https://pinstripes.io/pricing" + "supports_function_calling": true, + "supports_reasoning": true }, "pinstripes/ps/minimax-m2.7": { - "max_tokens": 1000192, + "input_cost_per_token": 2.55e-07, + "litellm_provider": "pinstripes", "max_input_tokens": 1000192, "max_output_tokens": 1000192, - "input_cost_per_token": 2.55e-07, - "output_cost_per_token": 5.5e-07, - "litellm_provider": "pinstripes", + "max_tokens": 1000192, "mode": "chat", - "supports_function_calling": true, + "output_cost_per_token": 5.5e-07, + "source": "https://pinstripes.io/pricing", "supports_assistant_prefill": true, - "supports_reasoning": false, - "source": "https://pinstripes.io/pricing" + "supports_function_calling": true, + "supports_reasoning": false }, "darkbloom/gemma-4-26b": { "input_cost_per_token": 3e-08, @@ -50138,63 +50144,63 @@ }, "xai/grok-4.20-0309-non-reasoning": { "cache_read_input_token_cost": 2e-07, + "cache_read_input_token_cost_above_200k_tokens": 4e-07, "input_cost_per_token": 1.25e-06, + "input_cost_per_token_above_200k_tokens": 2.5e-06, "litellm_provider": "xai", "max_input_tokens": 1000000, "max_output_tokens": 1000000, "max_tokens": 1000000, "mode": "chat", "output_cost_per_token": 2.5e-06, + "output_cost_per_token_above_200k_tokens": 5e-06, "source": "https://docs.x.ai/docs/models", "supports_function_calling": true, "supports_prompt_caching": true, + "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "input_cost_per_token_above_200k_tokens": 2.5e-06, - "output_cost_per_token_above_200k_tokens": 5e-06, - "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "supports_response_schema": true + "supports_web_search": true }, "xai/grok-4.20-multi-agent-0309": { "cache_read_input_token_cost": 2e-07, + "cache_read_input_token_cost_above_200k_tokens": 4e-07, "input_cost_per_token": 1.25e-06, + "input_cost_per_token_above_200k_tokens": 2.5e-06, "litellm_provider": "xai", "max_input_tokens": 1000000, "max_output_tokens": 1000000, "max_tokens": 1000000, "mode": "chat", "output_cost_per_token": 2.5e-06, + "output_cost_per_token_above_200k_tokens": 5e-06, "source": "https://docs.x.ai/docs/models", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, + "supports_response_schema": true, "supports_tool_choice": true, "supports_vision": true, - "supports_web_search": true, - "input_cost_per_token_above_200k_tokens": 2.5e-06, - "output_cost_per_token_above_200k_tokens": 5e-06, - "cache_read_input_token_cost_above_200k_tokens": 4e-07, - "supports_response_schema": true + "supports_web_search": true }, "xai/grok-build-0.1": { "cache_read_input_token_cost": 2e-07, + "cache_read_input_token_cost_above_200k_tokens": 4e-07, "input_cost_per_token": 1e-06, + "input_cost_per_token_above_200k_tokens": 2e-06, "litellm_provider": "xai", "max_input_tokens": 256000, "max_output_tokens": 256000, "max_tokens": 256000, "mode": "chat", "output_cost_per_token": 2e-06, + "output_cost_per_token_above_200k_tokens": 4e-06, "source": "https://docs.x.ai/docs/models", "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, - "supports_tool_choice": true, - "input_cost_per_token_above_200k_tokens": 2e-06, - "output_cost_per_token_above_200k_tokens": 4e-06, - "cache_read_input_token_cost_above_200k_tokens": 4e-07, "supports_response_schema": true, + "supports_tool_choice": true, "supports_vision": true }, "gpt-transcribe": { @@ -50270,12 +50276,12 @@ }, "source": "https://docs.claude.com/en/docs/about-claude/models/overview", "supports_adaptive_thinking": true, - "thinking_always_on": true, - "supports_mid_conversation_system": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, "supports_max_reasoning_effort": true, + "supports_mid_conversation_system": true, + "supports_native_structured_output": true, "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -50285,7 +50291,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true + "thinking_always_on": true }, "claude-mythos-preview": { "cache_creation_input_token_cost": 1.25e-05, @@ -50306,11 +50312,11 @@ }, "source": "https://docs.claude.com/en/docs/about-claude/models/overview", "supports_adaptive_thinking": true, - "thinking_always_on": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, "supports_max_reasoning_effort": true, + "supports_native_structured_output": true, "supports_output_config": true, "supports_pdf_input": true, "supports_prompt_caching": true, @@ -50320,7 +50326,7 @@ "supports_tool_choice": true, "supports_vision": true, "supports_xhigh_reasoning_effort": true, - "supports_native_structured_output": true + "thinking_always_on": true }, "gemini/gemini-robotics-er-2-streaming-preview": { "input_cost_per_audio_token": 2e-06, @@ -50485,6 +50491,14 @@ "supports_adaptive_thinking": true } }, + { + "name": "claude-legacy-thinking", + "pattern": "claude-[a-z]+-4[-._]6(?!\\d)", + "description": "Claude at version 4.6 exactly, in any id shape that contains claude--4-6 (dotted and underscored minors included, dated releases such as claude-sonnet-4-6-20260219 too). The 4.6 family is adaptive-thinking yet still accepts legacy thinking.type=enabled with budget_tokens, so the caller's hard budget cap is forwarded verbatim instead of being rewritten to an uncapped output_config.effort. The lookahead keeps two-digit minors such as 4-60 from matching. 4.7+ and 5+ majors reject the legacy shape and stay on the adaptive translation.", + "model_info": { + "supports_legacy_thinking": true + } + }, { "name": "claude-always-on-thinking", "pattern": "claude-(?:fable|mythos)-", @@ -50922,6 +50936,42 @@ "supports_tool_choice": true, "supports_vision": true }, + "moonshot/kimi-k2.7-code": { + "cache_read_input_token_cost": 1.9e-07, + "input_cost_per_token": 9.5e-07, + "litellm_provider": "moonshot", + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 4e-06, + "source": "https://platform.kimi.ai/docs/pricing/chat-k27-code", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_video_input": true, + "supports_vision": true + }, + "moonshot/kimi-k2.7-code-highspeed": { + "cache_read_input_token_cost": 3.8e-07, + "input_cost_per_token": 1.9e-06, + "litellm_provider": "moonshot", + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "max_tokens": 262144, + "mode": "chat", + "output_cost_per_token": 8e-06, + "source": "https://platform.kimi.ai/docs/pricing/chat-k27-code", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_video_input": true, + "supports_vision": true + }, "telnyx/moonshotai/Kimi-K2.6": { "litellm_provider": "telnyx", "max_input_tokens": 262144, diff --git a/model_prices_and_context_window.schema.json b/model_prices_and_context_window.schema.json index 84d08a3d904..ded0a373141 100644 --- a/model_prices_and_context_window.schema.json +++ b/model_prices_and_context_window.schema.json @@ -659,6 +659,9 @@ "supports_image_size": { "type": "boolean" }, + "supports_legacy_thinking": { + "type": "boolean" + }, "supports_low_reasoning_effort": { "type": "boolean" }, diff --git a/tests/test_litellm/llms/anthropic/experimental_pass_through/messages/test_reasoning_effort_translation.py b/tests/test_litellm/llms/anthropic/experimental_pass_through/messages/test_reasoning_effort_translation.py index f393a7b50b1..12ab536ed45 100644 --- a/tests/test_litellm/llms/anthropic/experimental_pass_through/messages/test_reasoning_effort_translation.py +++ b/tests/test_litellm/llms/anthropic/experimental_pass_through/messages/test_reasoning_effort_translation.py @@ -2,7 +2,6 @@ import pytest -import litellm from litellm.constants import ( DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET, DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET, @@ -17,7 +16,6 @@ from litellm.llms.bedrock.messages.invoke_transformations.anthropic_claude3_tran ) - @pytest.mark.parametrize( "reasoning_effort,expected_effort", [ @@ -258,19 +256,22 @@ def test_reasoning_effort_in_supported_params(): "model", [ "claude-sonnet-4-6", - "bedrock/invoke/us.anthropic.claude-sonnet-4-6", - "vertex_ai/claude-sonnet-4-6", "claude-opus-4-6", + "claude-sonnet-4-6-20260219", + "bedrock/invoke/us.anthropic.claude-sonnet-4-6", "bedrock/invoke/us.anthropic.claude-opus-4-6-v1:0", + "vertex_ai/claude-sonnet-4-6", "vertex_ai/claude-opus-4-6", + "azure_ai/claude-sonnet-4-6", ], ) -def test_legacy_thinking_high_budget_clamps_to_high_when_xhigh_unsupported( - local_model_cost_map, model -): - """Claude Code sends ``thinking.budget_tokens=31999``; Sonnet 4.6 and Opus 4.6 - have no ``xhigh`` tier, so the translator must emit ``high`` rather than the - provider-invalid ``xhigh`` (regression for issue #29282).""" +def test_legacy_thinking_budget_preserved_verbatim_on_46(local_model_cost_map, model): + """Regression for the passthrough silently dropping a caller's hard thinking + budget: the 4.6 family accepts ``thinking.type=enabled`` with ``budget_tokens`` + natively, so rewriting it to ``thinking.type=adaptive`` + ``output_config.effort`` + (which carries no ceiling) let reasoning run past the requested cap. The legacy + shape must be forwarded verbatim, in every 4.6 id shape including unmapped dated + releases resolved by the ``claude-legacy-thinking`` fallback rule.""" config = AnthropicMessagesConfig() optional_params = { "max_tokens": 1024, @@ -285,8 +286,8 @@ def test_legacy_thinking_high_budget_clamps_to_high_when_xhigh_unsupported( headers={}, ) - assert result.get("thinking") == {"type": "adaptive"} - assert result.get("output_config") == {"effort": "high"} + assert result.get("thinking") == {"type": "enabled", "budget_tokens": 31999} + assert "output_config" not in result def test_legacy_thinking_high_budget_keeps_xhigh_when_supported(): @@ -343,11 +344,44 @@ def test_legacy_thinking_translates_to_adaptive_for_opus_48( assert result.get("output_config") == {"effort": "xhigh"} +@pytest.mark.parametrize( + "model,expected_effort", + [ + ("claude-sonnet-5", "xhigh"), + ("claude-opus-5", "xhigh"), + ("claude-newfamily-6", "high"), + ], +) +def test_legacy_thinking_translates_to_adaptive_for_5_and_future_models( + local_model_cost_map, model, expected_effort +): + """The 5 families reject ``thinking.type=enabled``, so the adaptive translation + stays the safe default for every adaptive model not flagged + ``supports_legacy_thinking``, unmapped future ids included. An unmapped id + cannot prove ``xhigh`` support, so its high-budget bucket clamps to ``high``.""" + config = AnthropicMessagesConfig() + optional_params = { + "max_tokens": 1024, + "thinking": {"type": "enabled", "budget_tokens": 31999}, + } + + result = config.transform_anthropic_messages_request( + model=model, + messages=[{"role": "user", "content": "Hello"}], + anthropic_messages_optional_request_params=optional_params, + litellm_params={}, + headers={}, + ) + + assert result.get("thinking") == {"type": "adaptive"} + assert result.get("output_config") == {"effort": expected_effort} + + @pytest.mark.parametrize( "budget_tokens,expected_effort", [ - (DEFAULT_REASONING_EFFORT_XHIGH_THINKING_BUDGET * 2, "high"), - (DEFAULT_REASONING_EFFORT_XHIGH_THINKING_BUDGET, "high"), + (DEFAULT_REASONING_EFFORT_XHIGH_THINKING_BUDGET * 2, "xhigh"), + (DEFAULT_REASONING_EFFORT_XHIGH_THINKING_BUDGET, "xhigh"), (DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET, "high"), (DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET - 1, "medium"), (DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET, "medium"), @@ -355,7 +389,9 @@ def test_legacy_thinking_translates_to_adaptive_for_opus_48( (1, "low"), ], ) -def test_legacy_thinking_budget_buckets_on_sonnet_46(budget_tokens, expected_effort): +def test_legacy_thinking_budget_buckets_on_opus_48( + local_model_cost_map, budget_tokens, expected_effort +): config = AnthropicMessagesConfig() optional_params = { "max_tokens": 1024, @@ -363,7 +399,7 @@ def test_legacy_thinking_budget_buckets_on_sonnet_46(budget_tokens, expected_eff } result = config.transform_anthropic_messages_request( - model="claude-sonnet-4-6", + model="claude-opus-4-8", messages=[{"role": "user", "content": "Hello"}], anthropic_messages_optional_request_params=optional_params, litellm_params={}, @@ -373,7 +409,29 @@ def test_legacy_thinking_budget_buckets_on_sonnet_46(budget_tokens, expected_eff assert result.get("output_config") == {"effort": expected_effort} -def test_legacy_thinking_does_not_override_explicit_output_config(): +def test_legacy_thinking_does_not_override_explicit_output_config(local_model_cost_map): + config = AnthropicMessagesConfig() + optional_params = { + "max_tokens": 1024, + "thinking": {"type": "enabled", "budget_tokens": 31999}, + "output_config": {"effort": "low"}, + } + + result = config.transform_anthropic_messages_request( + model="claude-opus-4-8", + messages=[{"role": "user", "content": "Hello"}], + anthropic_messages_optional_request_params=optional_params, + litellm_params={}, + headers={}, + ) + + assert result.get("thinking") == {"type": "adaptive"} + assert result.get("output_config") == {"effort": "low"} + + +def test_legacy_thinking_with_explicit_output_config_untouched_on_46( + local_model_cost_map, +): config = AnthropicMessagesConfig() optional_params = { "max_tokens": 1024, @@ -389,6 +447,7 @@ def test_legacy_thinking_does_not_override_explicit_output_config(): headers={}, ) + assert result.get("thinking") == {"type": "enabled", "budget_tokens": 31999} assert result.get("output_config") == {"effort": "low"} diff --git a/tests/test_litellm/proxy/video_endpoints/test_endpoints.py b/tests/test_litellm/proxy/video_endpoints/test_endpoints.py index a959326817c..0dd163b7b94 100644 --- a/tests/test_litellm/proxy/video_endpoints/test_endpoints.py +++ b/tests/test_litellm/proxy/video_endpoints/test_endpoints.py @@ -372,6 +372,7 @@ async def test_content__model_encoded_id(harness): async def call_edit( harness: Harness, *, body: Dict[str, Any], headers=None, query=None ): + harness.read_body.return_value = dict(body) return await endpoints.video_edit( request=FakeRequest(headers=headers, query=query, raw_body=orjson.dumps(body)), fastapi_response=Response(), @@ -428,6 +429,27 @@ async def test_edit__missing_video_object_defaults_to_openai(harness): assert "video" not in data +@pytest.mark.asyncio +async def test_edit__bare_string_video_id_from_form_field(harness): + await call_edit(harness, body={"prompt": "brighter", "video": "video_plain"}) + + assert harness.processor_data() == { + "prompt": "brighter", + "video_id": "video_plain", + "custom_llm_provider": "openai", + } + + +@pytest.mark.asyncio +async def test_edit__json_string_video_reference_from_form_field(harness): + await call_edit( + harness, + body={"prompt": "brighter", "video": orjson.dumps({"id": "video_plain"}).decode()}, + ) + + assert harness.processor_data()["video_id"] == "video_plain" + + # =========================================================================== # # GET /v1/videos - video_list # # =========================================================================== # @@ -471,6 +493,7 @@ async def test_list__provider_from_header(harness): async def call_remix( harness: Harness, video_id: str, *, body, headers=None, query=None ): + harness.read_body.return_value = dict(body) return await endpoints.video_remix( video_id=video_id, request=FakeRequest(headers=headers, query=query, raw_body=orjson.dumps(body)), @@ -629,6 +652,7 @@ async def test_get_character__plain_id_defaults_openai_no_encode(harness): async def call_extension(harness: Harness, *, body, headers=None, query=None): + harness.read_body.return_value = dict(body) return await endpoints.video_extension( request=FakeRequest(headers=headers, query=query, raw_body=orjson.dumps(body)), fastapi_response=Response(), diff --git a/tests/test_litellm/proxy/video_endpoints/test_utils.py b/tests/test_litellm/proxy/video_endpoints/test_utils.py index efbaaff5f4b..9a2c208c075 100644 --- a/tests/test_litellm/proxy/video_endpoints/test_utils.py +++ b/tests/test_litellm/proxy/video_endpoints/test_utils.py @@ -1,8 +1,9 @@ """ Pure-logic contract tests for litellm/proxy/video_endpoints/utils.py -Three helpers the video proxy endpoints lean on: +Four helpers the video proxy endpoints lean on: - extract_model_from_target_model_names: first model from a comma string / list + - video_reference_to_id: normalize a video reference (dict / bare id / JSON string) to an id - get_custom_provider_from_data: provider precedence (top-level > extra_body) - encode_character_id_in_response: re-encode a response id in place @@ -20,6 +21,7 @@ from litellm.proxy.video_endpoints.utils import ( encode_character_id_in_response, extract_model_from_target_model_names, get_custom_provider_from_data, + video_reference_to_id, ) from litellm.types.videos.utils import ( decode_character_id_with_provider, @@ -53,6 +55,31 @@ def test_extract_model__non_str_non_list_is_none(value): assert extract_model_from_target_model_names(value) is None +# =========================================================================== # +# video_reference_to_id +# =========================================================================== # + + +@pytest.mark.parametrize( + "video_ref,expected", + [ + ({"id": "video_123"}, "video_123"), # dict reference -> its id + ({"id": ""}, ""), # dict with empty id + ({}, ""), # dict missing id -> default empty + ({"other": "x"}, ""), # dict without id key + ("video_123", "video_123"), # bare id string (not valid JSON) -> itself + ('{"id": "video_9"}', "video_9"), # JSON-encoded dict -> its id + ('{"other": 1}', ""), # JSON-encoded dict without id -> empty + ("[1, 2]", "[1, 2]"), # JSON parses to non-dict -> original string + (None, ""), # non-str, non-dict + (123, ""), # non-str, non-dict + (["video_123"], ""), # list is neither dict nor str + ], +) +def test_video_reference_to_id(video_ref, expected): + assert video_reference_to_id(video_ref) == expected + + # =========================================================================== # # get_custom_provider_from_data # =========================================================================== # diff --git a/tests/test_litellm/test_utils.py b/tests/test_litellm/test_utils.py index b418c237e7a..f946ec2fcf6 100644 --- a/tests/test_litellm/test_utils.py +++ b/tests/test_litellm/test_utils.py @@ -1001,6 +1001,7 @@ def test_aaamodel_prices_and_context_window_json_is_valid(): "supports_xhigh_reasoning_effort": {"type": "boolean"}, "supports_max_reasoning_effort": {"type": "boolean"}, "supports_adaptive_thinking": {"type": "boolean"}, + "supports_legacy_thinking": {"type": "boolean"}, "thinking_always_on": {"type": "boolean"}, "supports_mid_conversation_system": {"type": "boolean"}, "supports_sampling_params": {"type": "boolean"}, diff --git a/tests/test_litellm/test_video_generation.py b/tests/test_litellm/test_video_generation.py index fb167a8624e..dd06e290adf 100644 --- a/tests/test_litellm/test_video_generation.py +++ b/tests/test_litellm/test_video_generation.py @@ -2316,6 +2316,72 @@ def test_edit_and_extension_support_custom_provider_from_extra_body( assert captured_data["custom_llm_provider"] == "vertex_ai" +@pytest.mark.parametrize( + "handler_name, path, form", + [ + ( + "video_edit", + "/v1/videos/edits", + {"model": "my-video-model", "prompt": "brighter", "video": "video_123"}, + ), + ( + "video_extension", + "/v1/videos/extensions", + {"model": "my-video-model", "prompt": "continue", "seconds": "4", "video": "video_123"}, + ), + ], +) +@pytest.mark.asyncio +async def test_edit_and_extension_read_cached_body_after_auth_consumes_stream( + handler_name, path, form +): + from urllib.parse import urlencode + + from fastapi import Response + from starlette.requests import Request + + import litellm.proxy.video_endpoints.endpoints as endpoints + from litellm.proxy._types import ProxyException, UserAPIKeyAuth + from litellm.proxy.common_utils.http_parsing_utils import _read_request_body + + body = urlencode(form).encode() + stream = {"sent": False} + + async def receive(): + if stream["sent"]: + return {"type": "http.request", "body": b"", "more_body": False} + stream["sent"] = True + return {"type": "http.request", "body": body, "more_body": False} + + request = Request( + { + "type": "http", + "method": "POST", + "path": path, + "headers": [ + (b"content-type", b"application/x-www-form-urlencoded"), + (b"content-length", str(len(body)).encode()), + ], + "query_string": b"", + }, + receive, + ) + + await _read_request_body(request=request) + + handler = getattr(endpoints, handler_name) + with pytest.raises(ProxyException) as exc_info: + await handler( + request=request, + fastapi_response=Response(), + user_api_key_dict=UserAPIKeyAuth(api_key="sk-1234"), + ) + + message = str(exc_info.value) + assert "Stream consumed" not in message + assert "my-video-model" in message + + @pytest.mark.parametrize("endpoint", ["/v1/videos/edits", "/v1/videos/extensions"]) def test_edit_and_extension_route_with_encoded_video_ids( video_proxy_test_client, endpoint