From bfc7ae05bf78982b7cce01e5af37ad4b6e8574b5 Mon Sep 17 00:00:00 2001 From: unknown <> Date: Wed, 1 Jul 2026 14:20:26 +0000 Subject: [PATCH 01/11] feat: add claude-sonnet-5 pricing to model cost map Add Claude Sonnet 5 with introductory pricing (/0 per MTok, through Aug 31 2026) across Anthropic, Bedrock (global + regional), Vertex AI, Azure AI, and Snowflake providers. Includes cache pricing (1.25x write, 2x 1hr write, 0.1x read) and 128K output context. Fixes #31868 Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- litellm/constants.py | 1 + ...odel_prices_and_context_window_backup.json | 1836 +++++++++++------ model_prices_and_context_window.json | 1398 +++++++------ .../test_claude_sonnet_5_config.py | 163 ++ 4 files changed, 2137 insertions(+), 1261 deletions(-) create mode 100644 tests/test_litellm/test_claude_sonnet_5_config.py diff --git a/litellm/constants.py b/litellm/constants.py index aeb74a65839..1d17e7b6da6 100644 --- a/litellm/constants.py +++ b/litellm/constants.py @@ -1128,6 +1128,7 @@ BEDROCK_CONVERSE_MODELS = [ "anthropic.claude-opus-4-6-v1:0", "anthropic.claude-opus-4-6-v1", "anthropic.claude-sonnet-4-6", + "anthropic.claude-sonnet-5", "anthropic.claude-opus-4-1-20250805-v1:0", "anthropic.claude-opus-4-20250514-v1:0", "anthropic.claude-sonnet-4-20250514-v1:0", diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index 21132db93cb..15b16118a7b 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -1010,7 +1010,6 @@ "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, @@ -1042,7 +1041,6 @@ "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, @@ -1074,7 +1072,6 @@ "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, @@ -1106,7 +1103,6 @@ "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, @@ -1138,7 +1134,6 @@ "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, @@ -1170,7 +1165,6 @@ "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, @@ -1219,7 +1213,6 @@ "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, @@ -1253,7 +1246,6 @@ "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, @@ -1287,7 +1279,6 @@ "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, @@ -1321,7 +1312,6 @@ "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, @@ -1487,7 +1477,6 @@ "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, @@ -1521,7 +1510,6 @@ "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, @@ -1555,7 +1543,6 @@ "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, @@ -1589,7 +1576,6 @@ "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, @@ -1623,7 +1609,6 @@ "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, @@ -1688,7 +1673,6 @@ "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, @@ -1719,7 +1703,6 @@ "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, @@ -1750,7 +1733,6 @@ "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, @@ -1781,7 +1763,6 @@ "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, @@ -1812,7 +1793,6 @@ "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, @@ -1843,7 +1823,186 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "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, + "supports_native_structured_output": true, + "supports_output_config": true + }, + "anthropic.claude-sonnet-5": { "supports_adaptive_thinking": 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, + "input_cost_per_token": 2e-06, + "litellm_provider": "bedrock_converse", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1e-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_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, + "supports_native_structured_output": true, + "supports_output_config": true + }, + "global.anthropic.claude-sonnet-5": { + "supports_adaptive_thinking": 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, + "input_cost_per_token": 2e-06, + "litellm_provider": "bedrock_converse", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1e-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_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, + "supports_native_structured_output": true, + "supports_output_config": true + }, + "us.anthropic.claude-sonnet-5": { + "supports_adaptive_thinking": true, + "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, + "input_cost_per_token": 2.2e-06, + "litellm_provider": "bedrock_converse", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1.1e-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_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, + "supports_native_structured_output": true, + "supports_output_config": true + }, + "eu.anthropic.claude-sonnet-5": { + "supports_adaptive_thinking": true, + "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, + "input_cost_per_token": 2.2e-06, + "litellm_provider": "bedrock_converse", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1.1e-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_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, + "supports_native_structured_output": true, + "supports_output_config": true + }, + "au.anthropic.claude-sonnet-5": { + "supports_adaptive_thinking": true, + "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, + "input_cost_per_token": 2.2e-06, + "litellm_provider": "bedrock_converse", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1.1e-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_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, + "supports_native_structured_output": true, + "supports_output_config": true + }, + "jp.anthropic.claude-sonnet-5": { + "supports_adaptive_thinking": true, + "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, + "input_cost_per_token": 2.2e-06, + "litellm_provider": "bedrock_converse", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1.1e-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_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, @@ -2364,7 +2523,6 @@ "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, - "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, @@ -2394,7 +2552,6 @@ "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, - "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, @@ -2455,7 +2612,6 @@ "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, - "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, @@ -2523,6 +2679,34 @@ "max_tokens": 64000, "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, + "supports_output_config": true + }, + "azure_ai/claude-sonnet-5": { + "cache_creation_input_token_cost": 2.5e-06, + "cache_creation_input_token_cost_above_1hr": 4e-06, + "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 2e-06, + "litellm_provider": "azure_ai", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1e-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, "supports_assistant_prefill": true, "supports_computer_use": true, @@ -9132,17 +9316,16 @@ }, "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", - "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 + "source": "https://aws.amazon.com/bedrock/pricing/" }, "bedrock/us-east-1/moonshotai.kimi-k2-thinking": { "input_cost_per_token": 6e-07, @@ -9754,17 +9937,16 @@ }, "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", - "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 + "source": "https://aws.amazon.com/bedrock/pricing/" }, "bedrock/us-west-2/moonshotai.kimi-k2-thinking": { "input_cost_per_token": 6e-07, @@ -10274,6 +10456,35 @@ "supports_vision": true, "supports_output_config": true }, + "claude-sonnet-5": { + "cache_creation_input_token_cost": 2.5e-06, + "cache_creation_input_token_cost_above_1hr": 4e-06, + "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 2e-06, + "litellm_provider": "anthropic", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1e-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, + "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, + "supports_output_config": 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, @@ -16134,6 +16345,46 @@ "tpm": 8000000, "supports_image_size": false }, + "gemini-3-pro-image": { + "input_cost_per_image": 0.0011, + "input_cost_per_token": 2e-06, + "input_cost_per_token_batches": 1e-06, + "litellm_provider": "vertex_ai-language-models", + "max_input_tokens": 65536, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "image_generation", + "output_cost_per_image": 0.134, + "output_cost_per_image_token": 0.00012, + "output_cost_per_token": 1.2e-05, + "output_cost_per_token_batches": 6e-06, + "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-3-pro-image", + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/completions", + "/v1/batch" + ], + "supported_modalities": [ + "text", + "image" + ], + "supported_output_modalities": [ + "text", + "image" + ], + "supports_function_calling": false, + "supports_prompt_caching": true, + "supports_response_schema": 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" + }, "gemini-3-pro-image-preview": { "input_cost_per_image": 0.0011, "input_cost_per_token": 2e-06, @@ -16174,6 +16425,44 @@ }, "web_search_billing_unit": "per_query" }, + "gemini-3.1-flash-image": { + "input_cost_per_image": 0.00056, + "input_cost_per_token": 5e-07, + "litellm_provider": "vertex_ai-language-models", + "max_input_tokens": 65536, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "image_generation", + "output_cost_per_image": 0.0672, + "output_cost_per_image_token": 6e-05, + "output_cost_per_token": 3e-06, + "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models", + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/completions", + "/v1/batch" + ], + "supported_modalities": [ + "text", + "image" + ], + "supported_output_modalities": [ + "text", + "image" + ], + "supports_function_calling": false, + "supports_prompt_caching": true, + "supports_response_schema": 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" + }, "gemini-3.1-flash-image-preview": { "input_cost_per_image": 0.00056, "input_cost_per_token": 5e-07, @@ -17588,6 +17877,48 @@ }, "supports_image_size": false }, + "gemini/gemini-3-pro-image": { + "input_cost_per_image": 0.0011, + "input_cost_per_token": 2e-06, + "input_cost_per_token_batches": 1e-06, + "litellm_provider": "gemini", + "max_input_tokens": 65536, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "image_generation", + "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, + "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-3-pro-image", + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/completions", + "/v1/batch" + ], + "supported_modalities": [ + "text", + "image" + ], + "supported_output_modalities": [ + "text", + "image" + ], + "supports_function_calling": false, + "supports_prompt_caching": true, + "supports_response_schema": 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" + }, "gemini/gemini-3-pro-image-preview": { "input_cost_per_image": 0.0011, "input_cost_per_token": 2e-06, @@ -17630,6 +17961,47 @@ }, "web_search_billing_unit": "per_query" }, + "gemini/gemini-3.1-flash-image": { + "input_cost_per_token": 2.5e-07, + "input_cost_per_token_batches": 1.25e-07, + "litellm_provider": "gemini", + "max_input_tokens": 65536, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "image_generation", + "output_cost_per_image": 0.045, + "output_cost_per_image_token": 6e-05, + "output_cost_per_token": 1.5e-06, + "output_cost_per_token_batches": 7.5e-07, + "rpm": 1000, + "tpm": 4000000, + "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-3.1-flash-image", + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/completions", + "/v1/batch" + ], + "supported_modalities": [ + "text", + "image" + ], + "supported_output_modalities": [ + "text", + "image" + ], + "supports_function_calling": false, + "supports_prompt_caching": true, + "supports_response_schema": 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" + }, "gemini/gemini-3.1-flash-image-preview": { "input_cost_per_token": 2.5e-07, "input_cost_per_token_batches": 1.25e-07, @@ -18944,7 +19316,6 @@ "supported_endpoints": [ "/v1/chat/completions" ], - "supports_adaptive_thinking": true, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": true @@ -21696,8 +22067,8 @@ "output_cost_per_token_flex": 1.5e-05, "output_cost_per_token_batches": 1.5e-05, "output_cost_per_token_priority": 6e-05, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -21745,8 +22116,8 @@ "output_cost_per_token_flex": 1.5e-05, "output_cost_per_token_batches": 1.5e-05, "output_cost_per_token_priority": 6e-05, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -21790,8 +22161,8 @@ "output_cost_per_token_above_272k_tokens": 0.00027, "output_cost_per_token_flex": 9e-05, "output_cost_per_token_batches": 9e-05, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/responses", "/v1/batch" @@ -21835,8 +22206,8 @@ "output_cost_per_token_above_272k_tokens": 0.00027, "output_cost_per_token_flex": 9e-05, "output_cost_per_token_batches": 9e-05, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/responses", "/v1/batch" @@ -21884,8 +22255,8 @@ "output_cost_per_token_flex": 7.5e-06, "output_cost_per_token_batches": 7.5e-06, "output_cost_per_token_priority": 3e-05, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -21932,8 +22303,8 @@ "output_cost_per_token_flex": 7.5e-06, "output_cost_per_token_batches": 7.5e-06, "output_cost_per_token_priority": 3e-05, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -21973,8 +22344,8 @@ "output_cost_per_token_above_272k_tokens": 0.00027, "output_cost_per_token_flex": 9e-05, "output_cost_per_token_batches": 9e-05, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/responses", "/v1/batch" @@ -22017,8 +22388,8 @@ "output_cost_per_token_above_272k_tokens": 0.00027, "output_cost_per_token_flex": 9e-05, "output_cost_per_token_batches": 9e-05, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/responses", "/v1/batch" @@ -22054,7 +22425,7 @@ "input_cost_per_token_batches": 3.75e-07, "input_cost_per_token_priority": 1.5e-06, "litellm_provider": "openai", - "max_input_tokens": 272000, + "max_input_tokens": 1050000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", @@ -22062,8 +22433,8 @@ "output_cost_per_token_flex": 2.25e-06, "output_cost_per_token_batches": 2.25e-06, "output_cost_per_token_priority": 9e-06, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -22100,7 +22471,7 @@ "input_cost_per_token_batches": 3.75e-07, "input_cost_per_token_priority": 1.5e-06, "litellm_provider": "openai", - "max_input_tokens": 272000, + "max_input_tokens": 1050000, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", @@ -22108,8 +22479,8 @@ "output_cost_per_token_flex": 2.25e-06, "output_cost_per_token_batches": 2.25e-06, "output_cost_per_token_priority": 9e-06, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -22144,15 +22515,15 @@ "input_cost_per_token_flex": 1e-07, "input_cost_per_token_batches": 1e-07, "litellm_provider": "openai", - "max_input_tokens": 272000, + "max_input_tokens": 1050000, "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, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -22187,15 +22558,15 @@ "input_cost_per_token_flex": 1e-07, "input_cost_per_token_batches": 1e-07, "litellm_provider": "openai", - "max_input_tokens": 272000, + "max_input_tokens": 1050000, "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, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -22227,9 +22598,9 @@ "input_cost_per_token": 1.5e-05, "input_cost_per_token_batches": 7.5e-06, "litellm_provider": "openai", - "max_input_tokens": 128000, - "max_output_tokens": 272000, - "max_tokens": 272000, + "max_input_tokens": 400000, + "max_output_tokens": 128000, + "max_tokens": 128000, "mode": "responses", "output_cost_per_token": 0.00012, "output_cost_per_token_batches": 6e-05, @@ -22263,9 +22634,9 @@ "input_cost_per_token": 1.5e-05, "input_cost_per_token_batches": 7.5e-06, "litellm_provider": "openai", - "max_input_tokens": 128000, - "max_output_tokens": 272000, - "max_tokens": 272000, + "max_input_tokens": 400000, + "max_output_tokens": 128000, + "max_tokens": 128000, "mode": "responses", "output_cost_per_token": 0.00012, "output_cost_per_token_batches": 6e-05, @@ -24712,14 +25083,13 @@ }, "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, "supports_function_calling": true, - "supports_reasoning": true, "supports_system_messages": true, "supports_tool_choice": true, "source": "https://aws.amazon.com/bedrock/pricing/" @@ -28175,7 +28545,6 @@ "output_cost_per_token": 1.5e-05, "output_cost_per_token_above_200k_tokens": 2.25e-05, "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, @@ -28215,7 +28584,6 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, - "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, @@ -28277,7 +28645,6 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, - "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, @@ -29522,6 +29889,22 @@ "supports_reasoning": true, "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, + "litellm_provider": "openrouter", + "max_input_tokens": 202752, + "max_output_tokens": 65535, + "max_tokens": 65535, + "mode": "chat", + "source": "https://openrouter.ai/z-ai/glm-5.1", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, "openrouter/minimax/minimax-m2.1": { "input_cost_per_token": 2.7e-07, "output_cost_per_token": 1.2e-06, @@ -30200,7 +30583,6 @@ "supports_adaptive_thinking": true, "litellm_provider": "perplexity", "mode": "responses", - "supports_adaptive_thinking": true, "supports_web_search": true, "supports_reasoning": false, "supports_function_calling": true, @@ -30210,7 +30592,6 @@ "supports_adaptive_thinking": true, "litellm_provider": "perplexity", "mode": "responses", - "supports_adaptive_thinking": true, "supports_web_search": true, "supports_reasoning": false, "supports_function_calling": true, @@ -31126,7 +31507,7 @@ "supports_tool_choice": true }, "sambanova/Meta-Llama-3.2-1B-Instruct": { - "deprecation_date": "2025-06-25", + "deprecation_date": "2025-06-25", "input_cost_per_token": 4e-08, "litellm_provider": "sambanova", "max_input_tokens": 16384, @@ -31257,15 +31638,15 @@ "supports_vision": true, "source": "https://cloud.sambanova.ai/plans/pricing" }, - "snowflake/claude-3-5-sonnet": { + "snowflake/claude-3-5-sonnet": { "litellm_provider": "snowflake", "max_input_tokens": 200000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 0.000003, - "output_cost_per_token": 0.000015, - "cache_read_input_token_cost": 0.0000003, + "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, @@ -31273,14 +31654,14 @@ "supports_system_messages": true, "supports_response_schema": true }, - "snowflake/deepseek-r1": { + "snowflake/deepseek-r1": { "litellm_provider": "snowflake", "max_input_tokens": 128000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 0.00000135, - "output_cost_per_token": 0.0000054, + "input_cost_per_token": 1.35e-06, + "output_cost_per_token": 5.4e-06, "supports_reasoning": true, "supports_system_messages": true }, @@ -31339,8 +31720,8 @@ "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 0.0000012, - "output_cost_per_token": 0.0000012, + "input_cost_per_token": 1.2e-06, + "output_cost_per_token": 1.2e-06, "supports_function_calling": true, "supports_system_messages": true }, @@ -31350,8 +31731,8 @@ "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 0.00000072, - "output_cost_per_token": 0.00000072, + "input_cost_per_token": 7.2e-07, + "output_cost_per_token": 7.2e-07, "supports_function_calling": true, "supports_system_messages": true }, @@ -31361,8 +31742,8 @@ "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 0.00000024, - "output_cost_per_token": 0.00000024, + "input_cost_per_token": 2.4e-07, + "output_cost_per_token": 2.4e-07, "supports_system_messages": true }, "snowflake/llama3.2-1b": { @@ -31379,17 +31760,17 @@ "max_tokens": 8192, "mode": "chat" }, - "snowflake/llama3.3-70b": { + "snowflake/llama3.3-70b": { "max_tokens": 16384, "max_input_tokens": 128000, "max_output_tokens": 16384, - "input_cost_per_token": 0.00000072, - "output_cost_per_token": 0.00000072, + "input_cost_per_token": 7.2e-07, + "output_cost_per_token": 7.2e-07, "litellm_provider": "snowflake", "mode": "chat", "supports_function_calling": true, "supports_system_messages": true - }, + }, "snowflake/mistral-7b": { "litellm_provider": "snowflake", "max_input_tokens": 32000, @@ -31404,14 +31785,14 @@ "max_tokens": 8192, "mode": "chat" }, - "snowflake/mistral-large2": { + "snowflake/mistral-large2": { "litellm_provider": "snowflake", "max_input_tokens": 128000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 0.000002, - "output_cost_per_token": 0.000006, + "input_cost_per_token": 2e-06, + "output_cost_per_token": 6e-06, "supports_function_calling": true, "supports_system_messages": true, "supports_response_schema": true @@ -31451,17 +31832,17 @@ "max_tokens": 8192, "mode": "chat" }, - "snowflake/snowflake-llama-3.3-70b": { + "snowflake/snowflake-llama-3.3-70b": { "max_tokens": 16384, "max_input_tokens": 128000, "max_output_tokens": 16384, - "input_cost_per_token": 0.00000072, - "output_cost_per_token": 0.00000072, + "input_cost_per_token": 7.2e-07, + "output_cost_per_token": 7.2e-07, "litellm_provider": "snowflake", "mode": "chat", "supports_function_calling": true, "supports_system_messages": true - }, + }, "stability/sd3": { "litellm_provider": "stability", "mode": "image_generation", @@ -32713,7 +33094,7 @@ "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": 9.0e-06, + "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", @@ -33455,7 +33836,6 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.5e-05, - "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, @@ -34690,7 +35070,6 @@ "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, @@ -34720,7 +35099,6 @@ "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, @@ -34750,7 +35128,6 @@ "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, @@ -34781,7 +35158,6 @@ "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, @@ -34872,7 +35248,6 @@ "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, @@ -34903,7 +35278,6 @@ "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, @@ -34956,7 +35330,6 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, - "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, @@ -35265,6 +35638,21 @@ "tpm": 8000000, "supports_image_size": false }, + "vertex_ai/gemini-3-pro-image": { + "input_cost_per_image": 0.0011, + "input_cost_per_token": 2e-06, + "input_cost_per_token_batches": 1e-06, + "litellm_provider": "vertex_ai-language-models", + "max_input_tokens": 65536, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "image_generation", + "output_cost_per_image": 0.134, + "output_cost_per_image_token": 0.00012, + "output_cost_per_token": 1.2e-05, + "output_cost_per_token_batches": 6e-06, + "source": "https://docs.cloud.google.com/vertex-ai/generative-ai/docs/models/gemini/3-pro-image" + }, "vertex_ai/gemini-3-pro-image-preview": { "input_cost_per_image": 0.0011, "input_cost_per_token": 2e-06, @@ -35280,6 +35668,19 @@ "output_cost_per_token_batches": 6e-06, "source": "https://docs.cloud.google.com/vertex-ai/generative-ai/docs/models/gemini/3-pro-image" }, + "vertex_ai/gemini-3.1-flash-image": { + "input_cost_per_image": 0.00056, + "input_cost_per_token": 5e-07, + "litellm_provider": "vertex_ai-language-models", + "max_input_tokens": 65536, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "image_generation", + "output_cost_per_image": 0.0672, + "output_cost_per_image_token": 6e-05, + "output_cost_per_token": 3e-06, + "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models" + }, "vertex_ai/gemini-3.1-flash-image-preview": { "input_cost_per_image": 0.00056, "input_cost_per_token": 5e-07, @@ -37485,12 +37886,12 @@ }, "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, "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, @@ -37511,20 +37912,6 @@ "supports_tool_choice": true, "source": "https://aws.amazon.com/bedrock/pricing/" }, - "zai.glm-5": { - "input_cost_per_token": 1e-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_system_messages": true, - "supports_tool_choice": true - }, "zai/glm-5": { "cache_creation_input_token_cost": 0, "cache_read_input_token_cost": 2e-07, @@ -37540,6 +37927,21 @@ "supports_tool_choice": true, "source": "https://docs.z.ai/guides/overview/pricing" }, + "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", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "source": "https://docs.z.ai/guides/overview/pricing" + }, "zai/glm-5-code": { "cache_creation_input_token_cost": 0, "cache_read_input_token_cost": 3e-07, @@ -37570,6 +37972,21 @@ "supports_tool_choice": true, "source": "https://docs.z.ai/guides/overview/pricing" }, + "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", + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_tool_choice": true, + "source": "https://docs.z.ai/guides/overview/pricing" + }, "zai/glm-4.6": { "cache_creation_input_token_cost": 0, "cache_read_input_token_cost": 1.1e-07, @@ -42393,7 +42810,6 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, - "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, @@ -42411,6 +42827,64 @@ }, "supports_output_config": true }, + "vertex_ai/claude-sonnet-5": { + "cache_creation_input_token_cost": 2.5e-06, + "cache_creation_input_token_cost_above_1hr": 4e-06, + "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 2e-06, + "litellm_provider": "vertex_ai-anthropic_models", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1e-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, + "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, + "supports_output_config": true + }, + "vertex_ai/claude-sonnet-5@default": { + "cache_creation_input_token_cost": 2.5e-06, + "cache_creation_input_token_cost_above_1hr": 4e-06, + "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 2e-06, + "litellm_provider": "vertex_ai-anthropic_models", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1e-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, + "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, + "supports_output_config": true + }, "duckduckgo/search": { "litellm_provider": "duckduckgo", "mode": "search", @@ -42427,7 +42901,10 @@ "max_output_tokens": 32768, "max_tokens": 32768, "mode": "chat", - "supported_endpoints": ["/v1/chat/completions", "/v1/responses"], + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_reasoning": true, @@ -42442,7 +42919,10 @@ "max_output_tokens": 32768, "max_tokens": 32768, "mode": "chat", - "supported_endpoints": ["/v1/chat/completions", "/v1/responses"], + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_reasoning": true, @@ -42457,7 +42937,9 @@ "max_output_tokens": 65536, "max_tokens": 65536, "mode": "chat", - "supported_endpoints": ["/v1/chat/completions"], + "supported_endpoints": [ + "/v1/chat/completions" + ], "supports_function_calling": true, "supports_reasoning": true, "supports_response_schema": true, @@ -42471,7 +42953,9 @@ "max_output_tokens": 65536, "max_tokens": 65536, "mode": "chat", - "supported_endpoints": ["/v1/chat/completions"], + "supported_endpoints": [ + "/v1/chat/completions" + ], "supports_function_calling": true, "supports_reasoning": true, "supports_response_schema": true, @@ -42487,9 +42971,16 @@ "max_tokens": 128000, "mode": "responses", "use_openai_responses_path": true, - "supported_endpoints": ["/v1/responses"], - "supported_modalities": ["text", "image"], - "supported_output_modalities": ["text"], + "supported_endpoints": [ + "/v1/responses" + ], + "supported_modalities": [ + "text", + "image" + ], + "supported_output_modalities": [ + "text" + ], "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -42507,9 +42998,16 @@ "max_tokens": 128000, "mode": "responses", "use_openai_responses_path": true, - "supported_endpoints": ["/v1/responses"], - "supported_modalities": ["text", "image"], - "supported_output_modalities": ["text"], + "supported_endpoints": [ + "/v1/responses" + ], + "supported_modalities": [ + "text", + "image" + ], + "supported_output_modalities": [ + "text" + ], "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -42526,7 +43024,10 @@ "max_tokens": 256000, "mode": "chat", "use_openai_responses_path": true, - "supported_endpoints": ["/v1/chat/completions", "/v1/responses"], + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], "supports_function_calling": true, "supports_parallel_function_calling": false, "supports_reasoning": true, @@ -42542,7 +43043,10 @@ "max_tokens": 256000, "mode": "chat", "use_openai_responses_path": true, - "supported_endpoints": ["/v1/chat/completions", "/v1/responses"], + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], "supports_function_calling": true, "supports_parallel_function_calling": false, "supports_reasoning": true, @@ -42558,7 +43062,10 @@ "max_tokens": 128000, "mode": "chat", "use_openai_responses_path": true, - "supported_endpoints": ["/v1/chat/completions", "/v1/responses"], + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], "supports_function_calling": true, "supports_parallel_function_calling": false, "supports_reasoning": true, @@ -42717,20 +43224,6 @@ } ] }, - "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", - "supports_function_calling": true, - "supports_reasoning": true, - "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" - }, "bedrock/us-east-1/zai.glm-5": { "input_cost_per_token": 1e-06, "output_cost_per_token": 3.2e-06, @@ -42759,45 +43252,6 @@ "supports_tool_choice": true, "source": "https://aws.amazon.com/bedrock/pricing/" }, - "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", - "supports_function_calling": true, - "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" - }, - "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", - "supports_function_calling": true, - "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" - }, - "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", - "supports_function_calling": true, - "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" - }, "bedrock/us-gov-east-1/anthropic.claude-haiku-4-5-20251001-v1:0": { "cache_creation_input_token_cost": 1.5e-06, "cache_creation_input_token_cost_above_1hr": 2.4e-06, @@ -42844,362 +43298,382 @@ "supports_native_structured_output": true, "supports_pdf_input": true }, - "snowflake/claude-sonnet-4-5": { - "max_tokens": 16384, - "max_input_tokens": 200000, - "max_output_tokens": 16384, - "input_cost_per_token": 0.000003, - "output_cost_per_token": 0.000015, - "cache_read_input_token_cost": 0.0000003, - "litellm_provider": "snowflake", - "mode": "chat", - "supports_function_calling": true, - "supports_vision": true, - "supports_prompt_caching": true, - "supports_system_messages": true, - "supports_response_schema": true - }, - "snowflake/claude-sonnet-4-6": { - "max_tokens": 16384, - "max_input_tokens": 200000, - "max_output_tokens": 16384, - "input_cost_per_token": 0.000003, - "output_cost_per_token": 0.000015, - "cache_read_input_token_cost": 0.0000003, - "litellm_provider": "snowflake", - "mode": "chat", - "supports_function_calling": true, - "supports_vision": true, - "supports_prompt_caching": true, - "supports_system_messages": true, - "supports_response_schema": true - }, - "snowflake/claude-4-sonnet": { - "max_tokens": 16384, - "max_input_tokens": 200000, - "max_output_tokens": 16384, - "input_cost_per_token": 0.000003, - "output_cost_per_token": 0.000015, - "cache_read_input_token_cost": 0.0000003, - "litellm_provider": "snowflake", - "mode": "chat", - "supports_function_calling": true, - "supports_vision": true, - "supports_prompt_caching": true, - "supports_system_messages": true, - "supports_response_schema": true - }, - "snowflake/claude-4-opus": { - "max_tokens": 16384, - "max_input_tokens": 200000, - "max_output_tokens": 16384, - "input_cost_per_token": 0.000005, - "output_cost_per_token": 0.000025, - "cache_read_input_token_cost": 0.0000005, - "litellm_provider": "snowflake", - "mode": "chat", - "supports_function_calling": true, - "supports_vision": true, - "supports_prompt_caching": true, - "supports_system_messages": true, - "supports_reasoning": true, - "supports_response_schema": true - }, - "snowflake/claude-haiku-4-5": { - "max_tokens": 16384, - "max_input_tokens": 200000, - "max_output_tokens": 16384, - "input_cost_per_token": 0.000001, - "output_cost_per_token": 0.000005, - "cache_read_input_token_cost": 0.0000001, - "litellm_provider": "snowflake", - "mode": "chat", - "supports_function_calling": true, - "supports_vision": true, - "supports_prompt_caching": true, - "supports_system_messages": true, - "supports_response_schema": true - }, - "snowflake/claude-3-7-sonnet": { - "max_tokens": 16384, - "max_input_tokens": 200000, - "max_output_tokens": 16384, - "input_cost_per_token": 0.000003, - "output_cost_per_token": 0.000015, - "cache_read_input_token_cost": 0.0000003, - "litellm_provider": "snowflake", - "mode": "chat", - "supports_function_calling": true, - "supports_vision": true, - "supports_prompt_caching": true, - "supports_system_messages": true, - "supports_reasoning": true, - "supports_response_schema": true - }, - "snowflake/openai-gpt-4.1": { - "max_tokens": 16384, - "max_input_tokens": 300000, - "max_output_tokens": 16384, - "input_cost_per_token": 0.000002, - "output_cost_per_token": 0.000008, - "cache_read_input_token_cost": 0.0000005, - "litellm_provider": "snowflake", - "mode": "chat", - "supports_function_calling": true, - "supports_vision": true, - "supports_prompt_caching": true, - "supports_system_messages": true, - "supports_response_schema": true - }, - "snowflake/openai-gpt-5": { - "max_tokens": 16384, - "max_input_tokens": 300000, - "max_output_tokens": 16384, - "input_cost_per_token": 0.00000125, - "output_cost_per_token": 0.00001, - "cache_read_input_token_cost": 0.000000125, - "litellm_provider": "snowflake", - "mode": "chat", - "supports_function_calling": true, - "supports_vision": true, - "supports_prompt_caching": true, - "supports_system_messages": true, - "supports_reasoning": true, - "supports_response_schema": true - }, - "snowflake/openai-gpt-5-mini": { - "max_tokens": 16384, - "max_input_tokens": 1000000, - "max_output_tokens": 16384, - "input_cost_per_token": 0.0000003, - "output_cost_per_token": 0.0000012, - "litellm_provider": "snowflake", - "mode": "chat", - "supports_function_calling": true, - "supports_system_messages": true, - "supports_response_schema": true - }, - "snowflake/openai-gpt-5-nano": { - "max_tokens": 16384, - "max_input_tokens": 5000000, - "max_output_tokens": 16384, - "input_cost_per_token": 0.00000015, - "output_cost_per_token": 0.0000006, - "litellm_provider": "snowflake", - "mode": "chat", - "supports_function_calling": true, - "supports_system_messages": true, - "supports_response_schema": true - }, - "snowflake/llama4-maverick": { - "max_tokens": 16384, - "max_input_tokens": 128000, - "max_output_tokens": 16384, - "input_cost_per_token": 0.00000024, - "output_cost_per_token": 0.00000097, - "litellm_provider": "snowflake", - "mode": "chat", - "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": 0.00000007, - "output_cost_per_token": 0.0, - "litellm_provider": "snowflake", - "mode": "embedding" - }, - "snowflake/snowflake-arctic-embed-m-v2.0": { - "max_tokens": 8192, - "max_input_tokens": 8192, - "input_cost_per_token": 0.00000007, - "output_cost_per_token": 0.0, - "litellm_provider": "snowflake", - "mode": "embedding" - }, - "soniox/stt-async-v4": { - "litellm_provider": "soniox", - "max_output_tokens": 8000, - "max_tokens": 8000, - "input_cost_per_second": 0.0, - "output_cost_per_second": 0.0000277778, - "mode": "audio_transcription", - "source": "https://soniox.com/pricing", - "supported_endpoints": ["/v1/audio/transcriptions"], - "supports_audio_input": true - }, - "soniox/stt-async-v5": { - "litellm_provider": "soniox", - "max_output_tokens": 8000, - "max_tokens": 8000, - "input_cost_per_second": 0.0, - "output_cost_per_second": 0.0000277778, - "mode": "audio_transcription", - "source": "https://soniox.com/pricing", - "supported_endpoints": ["/v1/audio/transcriptions"], - "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, - "max_input_tokens": 262144, - "max_output_tokens": 262144, - "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" - }, - "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, - "max_input_tokens": 262144, - "max_output_tokens": 262144, - "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true, - "supports_prompt_caching": true, - "supports_system_messages": true, - "source": "https://serverless.tensormesh.ai/v1/models/openrouter" - }, - "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, - "max_input_tokens": 262144, - "max_output_tokens": 262144, - "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" - }, - "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, - "max_input_tokens": 202752, - "max_output_tokens": 202752, - "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" - }, - "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, - "max_input_tokens": 32768, - "max_output_tokens": 32768, - "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" - }, - "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, - "max_input_tokens": 32768, - "max_output_tokens": 32768, - "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" - }, - "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, - "max_input_tokens": 196608, - "max_output_tokens": 196608, - "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" - }, - "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, - "max_input_tokens": 32768, - "max_output_tokens": 32768, - "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" - }, - "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, - "max_input_tokens": 131072, - "max_output_tokens": 131072, - "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" - }, - "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, - "max_input_tokens": 131072, - "max_output_tokens": 131072, - "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" - } - , + "snowflake/claude-sonnet-4-5": { + "max_tokens": 16384, + "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", + "mode": "chat", + "supports_function_calling": true, + "supports_vision": true, + "supports_prompt_caching": true, + "supports_system_messages": true, + "supports_response_schema": true + }, + "snowflake/claude-sonnet-4-6": { + "supports_adaptive_thinking": true, + "max_tokens": 16384, + "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", + "mode": "chat", + "supports_function_calling": true, + "supports_vision": true, + "supports_prompt_caching": true, + "supports_system_messages": true, + "supports_response_schema": true + }, + "snowflake/claude-sonnet-5": { + "supports_adaptive_thinking": true, + "max_tokens": 16384, + "max_input_tokens": 200000, + "max_output_tokens": 16384, + "input_cost_per_token": 2e-06, + "output_cost_per_token": 1e-05, + "cache_read_input_token_cost": 2e-07, + "litellm_provider": "snowflake", + "mode": "chat", + "supports_function_calling": true, + "supports_vision": true, + "supports_prompt_caching": true, + "supports_system_messages": true, + "supports_response_schema": true + }, + "snowflake/claude-4-sonnet": { + "max_tokens": 16384, + "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", + "mode": "chat", + "supports_function_calling": true, + "supports_vision": true, + "supports_prompt_caching": true, + "supports_system_messages": true, + "supports_response_schema": true + }, + "snowflake/claude-4-opus": { + "max_tokens": 16384, + "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", + "mode": "chat", + "supports_function_calling": true, + "supports_vision": true, + "supports_prompt_caching": true, + "supports_system_messages": true, + "supports_reasoning": true, + "supports_response_schema": true + }, + "snowflake/claude-haiku-4-5": { + "max_tokens": 16384, + "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", + "mode": "chat", + "supports_function_calling": true, + "supports_vision": true, + "supports_prompt_caching": true, + "supports_system_messages": true, + "supports_response_schema": true + }, + "snowflake/claude-3-7-sonnet": { + "max_tokens": 16384, + "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", + "mode": "chat", + "supports_function_calling": true, + "supports_vision": true, + "supports_prompt_caching": true, + "supports_system_messages": true, + "supports_reasoning": true, + "supports_response_schema": true + }, + "snowflake/openai-gpt-4.1": { + "max_tokens": 16384, + "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", + "mode": "chat", + "supports_function_calling": true, + "supports_vision": true, + "supports_prompt_caching": true, + "supports_system_messages": true, + "supports_response_schema": true + }, + "snowflake/openai-gpt-5": { + "max_tokens": 16384, + "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", + "mode": "chat", + "supports_function_calling": true, + "supports_vision": true, + "supports_prompt_caching": true, + "supports_system_messages": true, + "supports_reasoning": true, + "supports_response_schema": true + }, + "snowflake/openai-gpt-5-mini": { + "max_tokens": 16384, + "max_input_tokens": 1000000, + "max_output_tokens": 16384, + "input_cost_per_token": 3e-07, + "output_cost_per_token": 1.2e-06, + "litellm_provider": "snowflake", + "mode": "chat", + "supports_function_calling": true, + "supports_system_messages": true, + "supports_response_schema": true + }, + "snowflake/openai-gpt-5-nano": { + "max_tokens": 16384, + "max_input_tokens": 5000000, + "max_output_tokens": 16384, + "input_cost_per_token": 1.5e-07, + "output_cost_per_token": 6e-07, + "litellm_provider": "snowflake", + "mode": "chat", + "supports_function_calling": true, + "supports_system_messages": true, + "supports_response_schema": true + }, + "snowflake/llama4-maverick": { + "max_tokens": 16384, + "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", + "mode": "chat", + "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" + }, + "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" + }, + "soniox/stt-async-v4": { + "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", + "source": "https://soniox.com/pricing", + "supported_endpoints": [ + "/v1/audio/transcriptions" + ], + "supports_audio_input": true + }, + "soniox/stt-async-v5": { + "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", + "source": "https://soniox.com/pricing", + "supported_endpoints": [ + "/v1/audio/transcriptions" + ], + "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, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "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" + }, + "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, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "supports_function_calling": true, + "supports_tool_choice": true, + "supports_response_schema": true, + "supports_prompt_caching": true, + "supports_system_messages": true, + "source": "https://serverless.tensormesh.ai/v1/models/openrouter" + }, + "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, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "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" + }, + "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, + "max_input_tokens": 202752, + "max_output_tokens": 202752, + "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" + }, + "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, + "max_input_tokens": 32768, + "max_output_tokens": 32768, + "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" + }, + "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, + "max_input_tokens": 32768, + "max_output_tokens": 32768, + "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" + }, + "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, + "max_input_tokens": 196608, + "max_output_tokens": 196608, + "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" + }, + "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, + "max_input_tokens": 32768, + "max_output_tokens": 32768, + "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" + }, + "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, + "max_input_tokens": 131072, + "max_output_tokens": 131072, + "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" + }, + "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, + "max_input_tokens": 131072, + "max_output_tokens": 131072, + "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" + }, "deepseek-v4-flash": { "cache_creation_input_token_cost": 0.0, "cache_read_input_token_cost": 2.8e-09, @@ -43275,6 +43749,109 @@ "supports_tool_choice": true, "supports_vision": false }, + "deepseek/deepseek-v4-pro": { + "cache_creation_input_token_cost": 0.0, + "cache_read_input_token_cost": 3.625e-09, + "input_cost_per_token": 4.35e-07, + "input_cost_per_token_cache_hit": 3.625e-09, + "litellm_provider": "deepseek", + "max_input_tokens": 1000000, + "max_output_tokens": 8192, + "max_tokens": 8192, + "mode": "chat", + "output_cost_per_token": 8.7e-07, + "source": "https://api-docs.deepseek.com/quick_start/pricing", + "supported_endpoints": [ + "/v1/chat/completions" + ], + "supports_assistant_prefill": true, + "supports_function_calling": true, + "supports_native_streaming": true, + "supports_parallel_function_calling": true, + "supports_prompt_caching": true, + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": false + }, + "pinstripes/ps/glm-4.5-air": { + "max_tokens": 128000, + "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", + "mode": "chat", + "supports_function_calling": true, + "supports_assistant_prefill": true, + "supports_reasoning": true, + "source": "https://pinstripes.io/pricing" + }, + "pinstripes/ps/qwen3.6-35b-a3b": { + "max_tokens": 131072, + "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", + "mode": "chat", + "supports_function_calling": true, + "supports_assistant_prefill": true, + "supports_reasoning": true, + "source": "https://pinstripes.io/pricing" + }, + "pinstripes/ps/qwen3-30b-a3b": { + "max_tokens": 131072, + "max_input_tokens": 131072, + "max_output_tokens": 131072, + "input_cost_per_token": 9e-08, + "output_cost_per_token": 2e-07, + "litellm_provider": "pinstripes", + "mode": "chat", + "supports_function_calling": true, + "supports_assistant_prefill": true, + "supports_reasoning": true, + "source": "https://pinstripes.io/pricing" + }, + "pinstripes/ps/qwen3-coder-30b-a3b": { + "max_tokens": 131072, + "max_input_tokens": 131072, + "max_output_tokens": 131072, + "input_cost_per_token": 3e-07, + "output_cost_per_token": 6e-07, + "litellm_provider": "pinstripes", + "mode": "chat", + "supports_function_calling": true, + "supports_assistant_prefill": true, + "supports_reasoning": false, + "source": "https://pinstripes.io/pricing" + }, + "pinstripes/ps/deepseek-v4-flash": { + "max_tokens": 163840, + "max_input_tokens": 163840, + "max_output_tokens": 163840, + "input_cost_per_token": 1e-07, + "output_cost_per_token": 2e-07, + "litellm_provider": "pinstripes", + "mode": "chat", + "supports_function_calling": true, + "supports_assistant_prefill": true, + "supports_reasoning": true, + "source": "https://pinstripes.io/pricing" + }, + "pinstripes/ps/minimax-m2.7": { + "max_tokens": 1000192, + "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", + "mode": "chat", + "supports_function_calling": true, + "supports_assistant_prefill": true, + "supports_reasoning": false, + "source": "https://pinstripes.io/pricing" + }, "darkbloom/gemma-4-26b": { "input_cost_per_token": 3e-08, "litellm_provider": "darkbloom", @@ -43309,142 +43886,39 @@ "supports_system_messages": true, "supports_tool_choice": true }, - "deepseek/deepseek-v4-pro": { - "cache_creation_input_token_cost": 0.0, - "cache_read_input_token_cost": 3.625e-09, - "input_cost_per_token": 4.35e-07, - "input_cost_per_token_cache_hit": 3.625e-09, - "litellm_provider": "deepseek", - "max_input_tokens": 1000000, - "max_output_tokens": 8192, - "max_tokens": 8192, - "mode": "chat", - "output_cost_per_token": 8.7e-07, - "source": "https://api-docs.deepseek.com/quick_start/pricing", - "supported_endpoints": [ - "/v1/chat/completions" - ], - "supports_assistant_prefill": true, - "supports_function_calling": true, - "supports_native_streaming": true, - "supports_parallel_function_calling": true, - "supports_prompt_caching": true, - "supports_response_schema": true, - "supports_system_messages": true, - "supports_tool_choice": true, - "supports_vision": false - }, - "pinstripes/ps/glm-4.5-air": { - "max_tokens": 128000, - "max_input_tokens": 128000, - "max_output_tokens": 128000, - "input_cost_per_token": 0.000000125, - "output_cost_per_token": 0.00000045, - "litellm_provider": "pinstripes", - "mode": "chat", - "supports_function_calling": true, - "supports_assistant_prefill": true, - "supports_reasoning": true, - "source": "https://pinstripes.io/pricing" - }, - "pinstripes/ps/qwen3.6-35b-a3b": { - "max_tokens": 131072, - "max_input_tokens": 131072, - "max_output_tokens": 131072, - "input_cost_per_token": 0.00000014, - "output_cost_per_token": 0.00000045, - "litellm_provider": "pinstripes", - "mode": "chat", - "supports_function_calling": true, - "supports_assistant_prefill": true, - "supports_reasoning": true, - "source": "https://pinstripes.io/pricing" - }, - "pinstripes/ps/qwen3-30b-a3b": { - "max_tokens": 131072, - "max_input_tokens": 131072, - "max_output_tokens": 131072, - "input_cost_per_token": 0.00000009, - "output_cost_per_token": 0.0000002, - "litellm_provider": "pinstripes", - "mode": "chat", - "supports_function_calling": true, - "supports_assistant_prefill": true, - "supports_reasoning": true, - "source": "https://pinstripes.io/pricing" - }, - "pinstripes/ps/qwen3-coder-30b-a3b": { - "max_tokens": 131072, - "max_input_tokens": 131072, - "max_output_tokens": 131072, - "input_cost_per_token": 0.0000003, - "output_cost_per_token": 0.0000006, - "litellm_provider": "pinstripes", - "mode": "chat", - "supports_function_calling": true, - "supports_assistant_prefill": true, - "supports_reasoning": false, - "source": "https://pinstripes.io/pricing" - }, - "pinstripes/ps/deepseek-v4-flash": { - "max_tokens": 163840, - "max_input_tokens": 163840, - "max_output_tokens": 163840, - "input_cost_per_token": 0.0000001, - "output_cost_per_token": 0.0000002, - "litellm_provider": "pinstripes", - "mode": "chat", - "supports_function_calling": true, - "supports_assistant_prefill": true, - "supports_reasoning": true, - "source": "https://pinstripes.io/pricing" - }, - "pinstripes/ps/minimax-m2.7": { - "max_tokens": 1000192, - "max_input_tokens": 1000192, - "max_output_tokens": 1000192, - "input_cost_per_token": 0.000000255, - "output_cost_per_token": 0.00000055, - "litellm_provider": "pinstripes", - "mode": "chat", - "supports_function_calling": true, - "supports_assistant_prefill": true, - "supports_reasoning": false, - "source": "https://pinstripes.io/pricing" - }, - "fallback_generalizations": { - "rules": [ - { - "name": "anthropic-claude-adaptive-thinking", - "pattern": "(?:opus|sonnet|haiku)[-._](?:4[-._](?:[6-9]|[1-9]\\d)(?!\\d)|(?:[5-9]|[1-9]\\d{1,})[-._]\\d{1,2}(?!\\d))", - "description": "Claude opus/sonnet/haiku at version 4.6 or higher: 4.6 through 4.99, then any 5.x, 6.x or later major. The minor is capped at two digits so an 8-digit date suffix such as claude-opus-4-20250514 is never read as a >= 4.6 minor. Turns on adaptive thinking for new families with no code change.", - "extends": "anthropic-claude", - "model_info": { - "supports_adaptive_thinking": true - } - }, - { - "name": "anthropic-claude", - "pattern": "^claude-[a-z]+-\\d+[-.]\\d+(?:-\\d{8})?$", - "description": "Any Claude family-major-minor id, optionally with an 8-digit date suffix, anchored to the whole name. Version-neutral fallback that gives an unmapped Claude provider routing and baseline capabilities; it carries no pricing, so cost stays on the standard unpriced behavior rather than a guessed number.", - "model_info": { - "litellm_provider": "anthropic", - "mode": "chat", - "max_input_tokens": 200000, - "max_output_tokens": 64000, - "max_tokens": 64000, - "supports_function_calling": true, - "supports_parallel_function_calling": true, - "supports_vision": true, - "supports_tool_choice": true, - "supports_assistant_prefill": true, - "supports_prompt_caching": true, - "supports_response_schema": true, - "supports_reasoning": true, - "supports_pdf_input": true, - "supports_system_messages": true - } - } - ] - } + "fallback_generalizations": { + "rules": [ + { + "name": "anthropic-claude-adaptive-thinking", + "pattern": "(?:opus|sonnet|haiku)[-._](?:4[-._](?:[6-9]|[1-9]\\d)(?!\\d)|(?:[5-9]|[1-9]\\d{1,})[-._]\\d{1,2}(?!\\d))", + "description": "Claude opus/sonnet/haiku at version 4.6 or higher: 4.6 through 4.99, then any 5.x, 6.x or later major. The minor is capped at two digits so an 8-digit date suffix such as claude-opus-4-20250514 is never read as a >= 4.6 minor. Turns on adaptive thinking for new families with no code change.", + "extends": "anthropic-claude", + "model_info": { + "supports_adaptive_thinking": true + } + }, + { + "name": "anthropic-claude", + "pattern": "^claude-[a-z]+-\\d+[-.]\\d+(?:-\\d{8})?$", + "description": "Any Claude family-major-minor id, optionally with an 8-digit date suffix, anchored to the whole name. Version-neutral fallback that gives an unmapped Claude provider routing and baseline capabilities; it carries no pricing, so cost stays on the standard unpriced behavior rather than a guessed number.", + "model_info": { + "litellm_provider": "anthropic", + "mode": "chat", + "max_input_tokens": 200000, + "max_output_tokens": 64000, + "max_tokens": 64000, + "supports_function_calling": true, + "supports_parallel_function_calling": true, + "supports_vision": true, + "supports_tool_choice": true, + "supports_assistant_prefill": true, + "supports_prompt_caching": true, + "supports_response_schema": true, + "supports_reasoning": true, + "supports_pdf_input": true, + "supports_system_messages": true + } + } + ] + } } diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index 73cefeb7c77..15b16118a7b 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -1010,7 +1010,6 @@ "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, @@ -1042,7 +1041,6 @@ "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, @@ -1074,7 +1072,6 @@ "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, @@ -1106,7 +1103,6 @@ "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, @@ -1138,7 +1134,6 @@ "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, @@ -1170,7 +1165,6 @@ "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, @@ -1219,7 +1213,6 @@ "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, @@ -1253,7 +1246,6 @@ "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, @@ -1287,7 +1279,6 @@ "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, @@ -1321,7 +1312,6 @@ "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, @@ -1487,7 +1477,6 @@ "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, @@ -1521,7 +1510,6 @@ "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, @@ -1555,7 +1543,6 @@ "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, @@ -1589,7 +1576,6 @@ "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, @@ -1623,7 +1609,6 @@ "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, @@ -1688,7 +1673,6 @@ "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, @@ -1719,7 +1703,6 @@ "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, @@ -1750,7 +1733,6 @@ "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, @@ -1781,7 +1763,6 @@ "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, @@ -1812,7 +1793,6 @@ "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, @@ -1843,7 +1823,186 @@ "search_context_size_low": 0.01, "search_context_size_medium": 0.01 }, + "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, + "supports_native_structured_output": true, + "supports_output_config": true + }, + "anthropic.claude-sonnet-5": { "supports_adaptive_thinking": 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, + "input_cost_per_token": 2e-06, + "litellm_provider": "bedrock_converse", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1e-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_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, + "supports_native_structured_output": true, + "supports_output_config": true + }, + "global.anthropic.claude-sonnet-5": { + "supports_adaptive_thinking": 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, + "input_cost_per_token": 2e-06, + "litellm_provider": "bedrock_converse", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1e-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_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, + "supports_native_structured_output": true, + "supports_output_config": true + }, + "us.anthropic.claude-sonnet-5": { + "supports_adaptive_thinking": true, + "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, + "input_cost_per_token": 2.2e-06, + "litellm_provider": "bedrock_converse", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1.1e-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_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, + "supports_native_structured_output": true, + "supports_output_config": true + }, + "eu.anthropic.claude-sonnet-5": { + "supports_adaptive_thinking": true, + "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, + "input_cost_per_token": 2.2e-06, + "litellm_provider": "bedrock_converse", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1.1e-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_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, + "supports_native_structured_output": true, + "supports_output_config": true + }, + "au.anthropic.claude-sonnet-5": { + "supports_adaptive_thinking": true, + "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, + "input_cost_per_token": 2.2e-06, + "litellm_provider": "bedrock_converse", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1.1e-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_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, + "supports_native_structured_output": true, + "supports_output_config": true + }, + "jp.anthropic.claude-sonnet-5": { + "supports_adaptive_thinking": true, + "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, + "input_cost_per_token": 2.2e-06, + "litellm_provider": "bedrock_converse", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1.1e-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_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, @@ -2364,7 +2523,6 @@ "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, - "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, @@ -2394,7 +2552,6 @@ "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, - "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, @@ -2455,7 +2612,6 @@ "cache_creation_input_token_cost": 6.25e-06, "cache_creation_input_token_cost_above_1hr": 1e-05, "cache_read_input_token_cost": 5e-07, - "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, @@ -2523,6 +2679,34 @@ "max_tokens": 64000, "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, + "supports_output_config": true + }, + "azure_ai/claude-sonnet-5": { + "cache_creation_input_token_cost": 2.5e-06, + "cache_creation_input_token_cost_above_1hr": 4e-06, + "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 2e-06, + "litellm_provider": "azure_ai", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1e-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, "supports_assistant_prefill": true, "supports_computer_use": true, @@ -9132,17 +9316,16 @@ }, "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", - "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 + "source": "https://aws.amazon.com/bedrock/pricing/" }, "bedrock/us-east-1/moonshotai.kimi-k2-thinking": { "input_cost_per_token": 6e-07, @@ -9754,17 +9937,16 @@ }, "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", - "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 + "source": "https://aws.amazon.com/bedrock/pricing/" }, "bedrock/us-west-2/moonshotai.kimi-k2-thinking": { "input_cost_per_token": 6e-07, @@ -10274,6 +10456,35 @@ "supports_vision": true, "supports_output_config": true }, + "claude-sonnet-5": { + "cache_creation_input_token_cost": 2.5e-06, + "cache_creation_input_token_cost_above_1hr": 4e-06, + "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 2e-06, + "litellm_provider": "anthropic", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1e-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, + "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, + "supports_output_config": 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, @@ -19105,7 +19316,6 @@ "supported_endpoints": [ "/v1/chat/completions" ], - "supports_adaptive_thinking": true, "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_vision": true @@ -21857,8 +22067,8 @@ "output_cost_per_token_flex": 1.5e-05, "output_cost_per_token_batches": 1.5e-05, "output_cost_per_token_priority": 6e-05, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -21906,8 +22116,8 @@ "output_cost_per_token_flex": 1.5e-05, "output_cost_per_token_batches": 1.5e-05, "output_cost_per_token_priority": 6e-05, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -21951,8 +22161,8 @@ "output_cost_per_token_above_272k_tokens": 0.00027, "output_cost_per_token_flex": 9e-05, "output_cost_per_token_batches": 9e-05, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/responses", "/v1/batch" @@ -21996,8 +22206,8 @@ "output_cost_per_token_above_272k_tokens": 0.00027, "output_cost_per_token_flex": 9e-05, "output_cost_per_token_batches": 9e-05, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/responses", "/v1/batch" @@ -22045,8 +22255,8 @@ "output_cost_per_token_flex": 7.5e-06, "output_cost_per_token_batches": 7.5e-06, "output_cost_per_token_priority": 3e-05, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -22093,8 +22303,8 @@ "output_cost_per_token_flex": 7.5e-06, "output_cost_per_token_batches": 7.5e-06, "output_cost_per_token_priority": 3e-05, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -22134,8 +22344,8 @@ "output_cost_per_token_above_272k_tokens": 0.00027, "output_cost_per_token_flex": 9e-05, "output_cost_per_token_batches": 9e-05, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/responses", "/v1/batch" @@ -22178,8 +22388,8 @@ "output_cost_per_token_above_272k_tokens": 0.00027, "output_cost_per_token_flex": 9e-05, "output_cost_per_token_batches": 9e-05, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/responses", "/v1/batch" @@ -22223,8 +22433,8 @@ "output_cost_per_token_flex": 2.25e-06, "output_cost_per_token_batches": 2.25e-06, "output_cost_per_token_priority": 9e-06, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -22269,8 +22479,8 @@ "output_cost_per_token_flex": 2.25e-06, "output_cost_per_token_batches": 2.25e-06, "output_cost_per_token_priority": 9e-06, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -22312,8 +22522,8 @@ "output_cost_per_token": 1.25e-06, "output_cost_per_token_flex": 6.25e-07, "output_cost_per_token_batches": 6.25e-07, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -22355,8 +22565,8 @@ "output_cost_per_token": 1.25e-06, "output_cost_per_token_flex": 6.25e-07, "output_cost_per_token_batches": 6.25e-07, - "regional_processing_uplift_multiplier_eu": 1.10, - "regional_processing_uplift_multiplier_us": 1.10, + "regional_processing_uplift_multiplier_eu": 1.1, + "regional_processing_uplift_multiplier_us": 1.1, "supported_endpoints": [ "/v1/chat/completions", "/v1/batch", @@ -24873,14 +25083,13 @@ }, "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, "supports_function_calling": true, - "supports_reasoning": true, "supports_system_messages": true, "supports_tool_choice": true, "source": "https://aws.amazon.com/bedrock/pricing/" @@ -28336,7 +28545,6 @@ "output_cost_per_token": 1.5e-05, "output_cost_per_token_above_200k_tokens": 2.25e-05, "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, @@ -28376,7 +28584,6 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, - "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, @@ -28438,7 +28645,6 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 2.5e-05, - "supports_adaptive_thinking": true, "supports_assistant_prefill": false, "supports_computer_use": true, "supports_function_calling": true, @@ -30377,7 +30583,6 @@ "supports_adaptive_thinking": true, "litellm_provider": "perplexity", "mode": "responses", - "supports_adaptive_thinking": true, "supports_web_search": true, "supports_reasoning": false, "supports_function_calling": true, @@ -30387,7 +30592,6 @@ "supports_adaptive_thinking": true, "litellm_provider": "perplexity", "mode": "responses", - "supports_adaptive_thinking": true, "supports_web_search": true, "supports_reasoning": false, "supports_function_calling": true, @@ -31434,15 +31638,15 @@ "supports_vision": true, "source": "https://cloud.sambanova.ai/plans/pricing" }, - "snowflake/claude-3-5-sonnet": { + "snowflake/claude-3-5-sonnet": { "litellm_provider": "snowflake", "max_input_tokens": 200000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 0.000003, - "output_cost_per_token": 0.000015, - "cache_read_input_token_cost": 0.0000003, + "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, @@ -31450,14 +31654,14 @@ "supports_system_messages": true, "supports_response_schema": true }, - "snowflake/deepseek-r1": { + "snowflake/deepseek-r1": { "litellm_provider": "snowflake", "max_input_tokens": 128000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 0.00000135, - "output_cost_per_token": 0.0000054, + "input_cost_per_token": 1.35e-06, + "output_cost_per_token": 5.4e-06, "supports_reasoning": true, "supports_system_messages": true }, @@ -31516,8 +31720,8 @@ "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 0.0000012, - "output_cost_per_token": 0.0000012, + "input_cost_per_token": 1.2e-06, + "output_cost_per_token": 1.2e-06, "supports_function_calling": true, "supports_system_messages": true }, @@ -31527,8 +31731,8 @@ "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 0.00000072, - "output_cost_per_token": 0.00000072, + "input_cost_per_token": 7.2e-07, + "output_cost_per_token": 7.2e-07, "supports_function_calling": true, "supports_system_messages": true }, @@ -31538,8 +31742,8 @@ "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 0.00000024, - "output_cost_per_token": 0.00000024, + "input_cost_per_token": 2.4e-07, + "output_cost_per_token": 2.4e-07, "supports_system_messages": true }, "snowflake/llama3.2-1b": { @@ -31556,17 +31760,17 @@ "max_tokens": 8192, "mode": "chat" }, - "snowflake/llama3.3-70b": { + "snowflake/llama3.3-70b": { "max_tokens": 16384, "max_input_tokens": 128000, "max_output_tokens": 16384, - "input_cost_per_token": 0.00000072, - "output_cost_per_token": 0.00000072, + "input_cost_per_token": 7.2e-07, + "output_cost_per_token": 7.2e-07, "litellm_provider": "snowflake", "mode": "chat", "supports_function_calling": true, "supports_system_messages": true - }, + }, "snowflake/mistral-7b": { "litellm_provider": "snowflake", "max_input_tokens": 32000, @@ -31581,14 +31785,14 @@ "max_tokens": 8192, "mode": "chat" }, - "snowflake/mistral-large2": { + "snowflake/mistral-large2": { "litellm_provider": "snowflake", "max_input_tokens": 128000, "max_output_tokens": 16384, "max_tokens": 16384, "mode": "chat", - "input_cost_per_token": 0.000002, - "output_cost_per_token": 0.000006, + "input_cost_per_token": 2e-06, + "output_cost_per_token": 6e-06, "supports_function_calling": true, "supports_system_messages": true, "supports_response_schema": true @@ -31628,17 +31832,17 @@ "max_tokens": 8192, "mode": "chat" }, - "snowflake/snowflake-llama-3.3-70b": { + "snowflake/snowflake-llama-3.3-70b": { "max_tokens": 16384, "max_input_tokens": 128000, "max_output_tokens": 16384, - "input_cost_per_token": 0.00000072, - "output_cost_per_token": 0.00000072, + "input_cost_per_token": 7.2e-07, + "output_cost_per_token": 7.2e-07, "litellm_provider": "snowflake", "mode": "chat", "supports_function_calling": true, "supports_system_messages": true - }, + }, "stability/sd3": { "litellm_provider": "stability", "mode": "image_generation", @@ -32890,7 +33094,7 @@ "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": 9.0e-06, + "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", @@ -33632,7 +33836,6 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 2.5e-05, - "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, @@ -34867,7 +35070,6 @@ "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, @@ -34897,7 +35099,6 @@ "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, @@ -34927,7 +35128,6 @@ "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, @@ -34958,7 +35158,6 @@ "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, @@ -35049,7 +35248,6 @@ "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, @@ -35080,7 +35278,6 @@ "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, @@ -35133,7 +35330,6 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, - "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, @@ -37690,12 +37886,12 @@ }, "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, "supports_function_calling": true, "supports_reasoning": true, "supports_system_messages": true, @@ -37716,20 +37912,6 @@ "supports_tool_choice": true, "source": "https://aws.amazon.com/bedrock/pricing/" }, - "zai.glm-5": { - "input_cost_per_token": 1e-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_system_messages": true, - "supports_tool_choice": true - }, "zai/glm-5": { "cache_creation_input_token_cost": 0, "cache_read_input_token_cost": 2e-07, @@ -42628,7 +42810,6 @@ "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 1.5e-05, - "supports_adaptive_thinking": true, "supports_assistant_prefill": true, "supports_computer_use": true, "supports_function_calling": true, @@ -42646,6 +42827,64 @@ }, "supports_output_config": true }, + "vertex_ai/claude-sonnet-5": { + "cache_creation_input_token_cost": 2.5e-06, + "cache_creation_input_token_cost_above_1hr": 4e-06, + "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 2e-06, + "litellm_provider": "vertex_ai-anthropic_models", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1e-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, + "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, + "supports_output_config": true + }, + "vertex_ai/claude-sonnet-5@default": { + "cache_creation_input_token_cost": 2.5e-06, + "cache_creation_input_token_cost_above_1hr": 4e-06, + "cache_read_input_token_cost": 2e-07, + "input_cost_per_token": 2e-06, + "litellm_provider": "vertex_ai-anthropic_models", + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 1e-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, + "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, + "supports_output_config": true + }, "duckduckgo/search": { "litellm_provider": "duckduckgo", "mode": "search", @@ -42662,7 +42901,10 @@ "max_output_tokens": 32768, "max_tokens": 32768, "mode": "chat", - "supported_endpoints": ["/v1/chat/completions", "/v1/responses"], + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_reasoning": true, @@ -42677,7 +42919,10 @@ "max_output_tokens": 32768, "max_tokens": 32768, "mode": "chat", - "supported_endpoints": ["/v1/chat/completions", "/v1/responses"], + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], "supports_function_calling": true, "supports_parallel_function_calling": true, "supports_reasoning": true, @@ -42692,7 +42937,9 @@ "max_output_tokens": 65536, "max_tokens": 65536, "mode": "chat", - "supported_endpoints": ["/v1/chat/completions"], + "supported_endpoints": [ + "/v1/chat/completions" + ], "supports_function_calling": true, "supports_reasoning": true, "supports_response_schema": true, @@ -42706,7 +42953,9 @@ "max_output_tokens": 65536, "max_tokens": 65536, "mode": "chat", - "supported_endpoints": ["/v1/chat/completions"], + "supported_endpoints": [ + "/v1/chat/completions" + ], "supports_function_calling": true, "supports_reasoning": true, "supports_response_schema": true, @@ -42722,9 +42971,16 @@ "max_tokens": 128000, "mode": "responses", "use_openai_responses_path": true, - "supported_endpoints": ["/v1/responses"], - "supported_modalities": ["text", "image"], - "supported_output_modalities": ["text"], + "supported_endpoints": [ + "/v1/responses" + ], + "supported_modalities": [ + "text", + "image" + ], + "supported_output_modalities": [ + "text" + ], "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -42742,9 +42998,16 @@ "max_tokens": 128000, "mode": "responses", "use_openai_responses_path": true, - "supported_endpoints": ["/v1/responses"], - "supported_modalities": ["text", "image"], - "supported_output_modalities": ["text"], + "supported_endpoints": [ + "/v1/responses" + ], + "supported_modalities": [ + "text", + "image" + ], + "supported_output_modalities": [ + "text" + ], "supports_function_calling": true, "supports_prompt_caching": true, "supports_reasoning": true, @@ -42761,7 +43024,10 @@ "max_tokens": 256000, "mode": "chat", "use_openai_responses_path": true, - "supported_endpoints": ["/v1/chat/completions", "/v1/responses"], + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], "supports_function_calling": true, "supports_parallel_function_calling": false, "supports_reasoning": true, @@ -42777,7 +43043,10 @@ "max_tokens": 256000, "mode": "chat", "use_openai_responses_path": true, - "supported_endpoints": ["/v1/chat/completions", "/v1/responses"], + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], "supports_function_calling": true, "supports_parallel_function_calling": false, "supports_reasoning": true, @@ -42793,7 +43062,10 @@ "max_tokens": 128000, "mode": "chat", "use_openai_responses_path": true, - "supported_endpoints": ["/v1/chat/completions", "/v1/responses"], + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], "supports_function_calling": true, "supports_parallel_function_calling": false, "supports_reasoning": true, @@ -42952,20 +43224,6 @@ } ] }, - "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", - "supports_function_calling": true, - "supports_reasoning": true, - "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" - }, "bedrock/us-east-1/zai.glm-5": { "input_cost_per_token": 1e-06, "output_cost_per_token": 3.2e-06, @@ -42994,45 +43252,6 @@ "supports_tool_choice": true, "source": "https://aws.amazon.com/bedrock/pricing/" }, - "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", - "supports_function_calling": true, - "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" - }, - "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", - "supports_function_calling": true, - "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" - }, - "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", - "supports_function_calling": true, - "supports_system_messages": true, - "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" - }, "bedrock/us-gov-east-1/anthropic.claude-haiku-4-5-20251001-v1:0": { "cache_creation_input_token_cost": 1.5e-06, "cache_creation_input_token_cost_above_1hr": 2.4e-06, @@ -43079,363 +43298,382 @@ "supports_native_structured_output": true, "supports_pdf_input": true }, - "snowflake/claude-sonnet-4-5": { - "max_tokens": 16384, - "max_input_tokens": 200000, - "max_output_tokens": 16384, - "input_cost_per_token": 0.000003, - "output_cost_per_token": 0.000015, - "cache_read_input_token_cost": 0.0000003, - "litellm_provider": "snowflake", - "mode": "chat", - "supports_function_calling": true, - "supports_vision": true, - "supports_prompt_caching": true, - "supports_system_messages": true, - "supports_response_schema": true - }, - "snowflake/claude-sonnet-4-6": { - "supports_adaptive_thinking": true, - "max_tokens": 16384, - "max_input_tokens": 200000, - "max_output_tokens": 16384, - "input_cost_per_token": 0.000003, - "output_cost_per_token": 0.000015, - "cache_read_input_token_cost": 0.0000003, - "litellm_provider": "snowflake", - "mode": "chat", - "supports_function_calling": true, - "supports_vision": true, - "supports_prompt_caching": true, - "supports_system_messages": true, - "supports_response_schema": true - }, - "snowflake/claude-4-sonnet": { - "max_tokens": 16384, - "max_input_tokens": 200000, - "max_output_tokens": 16384, - "input_cost_per_token": 0.000003, - "output_cost_per_token": 0.000015, - "cache_read_input_token_cost": 0.0000003, - "litellm_provider": "snowflake", - "mode": "chat", - "supports_function_calling": true, - "supports_vision": true, - "supports_prompt_caching": true, - "supports_system_messages": true, - "supports_response_schema": true - }, - "snowflake/claude-4-opus": { - "max_tokens": 16384, - "max_input_tokens": 200000, - "max_output_tokens": 16384, - "input_cost_per_token": 0.000005, - "output_cost_per_token": 0.000025, - "cache_read_input_token_cost": 0.0000005, - "litellm_provider": "snowflake", - "mode": "chat", - "supports_function_calling": true, - "supports_vision": true, - "supports_prompt_caching": true, - "supports_system_messages": true, - "supports_reasoning": true, - "supports_response_schema": true - }, - "snowflake/claude-haiku-4-5": { - "max_tokens": 16384, - "max_input_tokens": 200000, - "max_output_tokens": 16384, - "input_cost_per_token": 0.000001, - "output_cost_per_token": 0.000005, - "cache_read_input_token_cost": 0.0000001, - "litellm_provider": "snowflake", - "mode": "chat", - "supports_function_calling": true, - "supports_vision": true, - "supports_prompt_caching": true, - "supports_system_messages": true, - "supports_response_schema": true - }, - "snowflake/claude-3-7-sonnet": { - "max_tokens": 16384, - "max_input_tokens": 200000, - "max_output_tokens": 16384, - "input_cost_per_token": 0.000003, - "output_cost_per_token": 0.000015, - "cache_read_input_token_cost": 0.0000003, - "litellm_provider": "snowflake", - "mode": "chat", - "supports_function_calling": true, - "supports_vision": true, - "supports_prompt_caching": true, - "supports_system_messages": true, - "supports_reasoning": true, - "supports_response_schema": true - }, - "snowflake/openai-gpt-4.1": { - "max_tokens": 16384, - "max_input_tokens": 300000, - "max_output_tokens": 16384, - "input_cost_per_token": 0.000002, - "output_cost_per_token": 0.000008, - "cache_read_input_token_cost": 0.0000005, - "litellm_provider": "snowflake", - "mode": "chat", - "supports_function_calling": true, - "supports_vision": true, - "supports_prompt_caching": true, - "supports_system_messages": true, - "supports_response_schema": true - }, - "snowflake/openai-gpt-5": { - "max_tokens": 16384, - "max_input_tokens": 300000, - "max_output_tokens": 16384, - "input_cost_per_token": 0.00000125, - "output_cost_per_token": 0.00001, - "cache_read_input_token_cost": 0.000000125, - "litellm_provider": "snowflake", - "mode": "chat", - "supports_function_calling": true, - "supports_vision": true, - "supports_prompt_caching": true, - "supports_system_messages": true, - "supports_reasoning": true, - "supports_response_schema": true - }, - "snowflake/openai-gpt-5-mini": { - "max_tokens": 16384, - "max_input_tokens": 1000000, - "max_output_tokens": 16384, - "input_cost_per_token": 0.0000003, - "output_cost_per_token": 0.0000012, - "litellm_provider": "snowflake", - "mode": "chat", - "supports_function_calling": true, - "supports_system_messages": true, - "supports_response_schema": true - }, - "snowflake/openai-gpt-5-nano": { - "max_tokens": 16384, - "max_input_tokens": 5000000, - "max_output_tokens": 16384, - "input_cost_per_token": 0.00000015, - "output_cost_per_token": 0.0000006, - "litellm_provider": "snowflake", - "mode": "chat", - "supports_function_calling": true, - "supports_system_messages": true, - "supports_response_schema": true - }, - "snowflake/llama4-maverick": { - "max_tokens": 16384, - "max_input_tokens": 128000, - "max_output_tokens": 16384, - "input_cost_per_token": 0.00000024, - "output_cost_per_token": 0.00000097, - "litellm_provider": "snowflake", - "mode": "chat", - "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": 0.00000007, - "output_cost_per_token": 0.0, - "litellm_provider": "snowflake", - "mode": "embedding" - }, - "snowflake/snowflake-arctic-embed-m-v2.0": { - "max_tokens": 8192, - "max_input_tokens": 8192, - "input_cost_per_token": 0.00000007, - "output_cost_per_token": 0.0, - "litellm_provider": "snowflake", - "mode": "embedding" - }, - "soniox/stt-async-v4": { - "litellm_provider": "soniox", - "max_output_tokens": 8000, - "max_tokens": 8000, - "input_cost_per_second": 0.0, - "output_cost_per_second": 0.0000277778, - "mode": "audio_transcription", - "source": "https://soniox.com/pricing", - "supported_endpoints": ["/v1/audio/transcriptions"], - "supports_audio_input": true - }, - "soniox/stt-async-v5": { - "litellm_provider": "soniox", - "max_output_tokens": 8000, - "max_tokens": 8000, - "input_cost_per_second": 0.0, - "output_cost_per_second": 0.0000277778, - "mode": "audio_transcription", - "source": "https://soniox.com/pricing", - "supported_endpoints": ["/v1/audio/transcriptions"], - "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, - "max_input_tokens": 262144, - "max_output_tokens": 262144, - "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" - }, - "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, - "max_input_tokens": 262144, - "max_output_tokens": 262144, - "supports_function_calling": true, - "supports_tool_choice": true, - "supports_response_schema": true, - "supports_prompt_caching": true, - "supports_system_messages": true, - "source": "https://serverless.tensormesh.ai/v1/models/openrouter" - }, - "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, - "max_input_tokens": 262144, - "max_output_tokens": 262144, - "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" - }, - "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, - "max_input_tokens": 202752, - "max_output_tokens": 202752, - "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" - }, - "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, - "max_input_tokens": 32768, - "max_output_tokens": 32768, - "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" - }, - "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, - "max_input_tokens": 32768, - "max_output_tokens": 32768, - "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" - }, - "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, - "max_input_tokens": 196608, - "max_output_tokens": 196608, - "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" - }, - "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, - "max_input_tokens": 32768, - "max_output_tokens": 32768, - "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" - }, - "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, - "max_input_tokens": 131072, - "max_output_tokens": 131072, - "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" - }, - "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, - "max_input_tokens": 131072, - "max_output_tokens": 131072, - "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" - } - , + "snowflake/claude-sonnet-4-5": { + "max_tokens": 16384, + "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", + "mode": "chat", + "supports_function_calling": true, + "supports_vision": true, + "supports_prompt_caching": true, + "supports_system_messages": true, + "supports_response_schema": true + }, + "snowflake/claude-sonnet-4-6": { + "supports_adaptive_thinking": true, + "max_tokens": 16384, + "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", + "mode": "chat", + "supports_function_calling": true, + "supports_vision": true, + "supports_prompt_caching": true, + "supports_system_messages": true, + "supports_response_schema": true + }, + "snowflake/claude-sonnet-5": { + "supports_adaptive_thinking": true, + "max_tokens": 16384, + "max_input_tokens": 200000, + "max_output_tokens": 16384, + "input_cost_per_token": 2e-06, + "output_cost_per_token": 1e-05, + "cache_read_input_token_cost": 2e-07, + "litellm_provider": "snowflake", + "mode": "chat", + "supports_function_calling": true, + "supports_vision": true, + "supports_prompt_caching": true, + "supports_system_messages": true, + "supports_response_schema": true + }, + "snowflake/claude-4-sonnet": { + "max_tokens": 16384, + "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", + "mode": "chat", + "supports_function_calling": true, + "supports_vision": true, + "supports_prompt_caching": true, + "supports_system_messages": true, + "supports_response_schema": true + }, + "snowflake/claude-4-opus": { + "max_tokens": 16384, + "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", + "mode": "chat", + "supports_function_calling": true, + "supports_vision": true, + "supports_prompt_caching": true, + "supports_system_messages": true, + "supports_reasoning": true, + "supports_response_schema": true + }, + "snowflake/claude-haiku-4-5": { + "max_tokens": 16384, + "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", + "mode": "chat", + "supports_function_calling": true, + "supports_vision": true, + "supports_prompt_caching": true, + "supports_system_messages": true, + "supports_response_schema": true + }, + "snowflake/claude-3-7-sonnet": { + "max_tokens": 16384, + "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", + "mode": "chat", + "supports_function_calling": true, + "supports_vision": true, + "supports_prompt_caching": true, + "supports_system_messages": true, + "supports_reasoning": true, + "supports_response_schema": true + }, + "snowflake/openai-gpt-4.1": { + "max_tokens": 16384, + "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", + "mode": "chat", + "supports_function_calling": true, + "supports_vision": true, + "supports_prompt_caching": true, + "supports_system_messages": true, + "supports_response_schema": true + }, + "snowflake/openai-gpt-5": { + "max_tokens": 16384, + "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", + "mode": "chat", + "supports_function_calling": true, + "supports_vision": true, + "supports_prompt_caching": true, + "supports_system_messages": true, + "supports_reasoning": true, + "supports_response_schema": true + }, + "snowflake/openai-gpt-5-mini": { + "max_tokens": 16384, + "max_input_tokens": 1000000, + "max_output_tokens": 16384, + "input_cost_per_token": 3e-07, + "output_cost_per_token": 1.2e-06, + "litellm_provider": "snowflake", + "mode": "chat", + "supports_function_calling": true, + "supports_system_messages": true, + "supports_response_schema": true + }, + "snowflake/openai-gpt-5-nano": { + "max_tokens": 16384, + "max_input_tokens": 5000000, + "max_output_tokens": 16384, + "input_cost_per_token": 1.5e-07, + "output_cost_per_token": 6e-07, + "litellm_provider": "snowflake", + "mode": "chat", + "supports_function_calling": true, + "supports_system_messages": true, + "supports_response_schema": true + }, + "snowflake/llama4-maverick": { + "max_tokens": 16384, + "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", + "mode": "chat", + "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" + }, + "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" + }, + "soniox/stt-async-v4": { + "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", + "source": "https://soniox.com/pricing", + "supported_endpoints": [ + "/v1/audio/transcriptions" + ], + "supports_audio_input": true + }, + "soniox/stt-async-v5": { + "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", + "source": "https://soniox.com/pricing", + "supported_endpoints": [ + "/v1/audio/transcriptions" + ], + "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, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "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" + }, + "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, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "supports_function_calling": true, + "supports_tool_choice": true, + "supports_response_schema": true, + "supports_prompt_caching": true, + "supports_system_messages": true, + "source": "https://serverless.tensormesh.ai/v1/models/openrouter" + }, + "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, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "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" + }, + "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, + "max_input_tokens": 202752, + "max_output_tokens": 202752, + "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" + }, + "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, + "max_input_tokens": 32768, + "max_output_tokens": 32768, + "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" + }, + "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, + "max_input_tokens": 32768, + "max_output_tokens": 32768, + "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" + }, + "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, + "max_input_tokens": 196608, + "max_output_tokens": 196608, + "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" + }, + "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, + "max_input_tokens": 32768, + "max_output_tokens": 32768, + "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" + }, + "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, + "max_input_tokens": 131072, + "max_output_tokens": 131072, + "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" + }, + "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, + "max_input_tokens": 131072, + "max_output_tokens": 131072, + "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" + }, "deepseek-v4-flash": { "cache_creation_input_token_cost": 0.0, "cache_read_input_token_cost": 2.8e-09, @@ -43540,8 +43778,8 @@ "max_tokens": 128000, "max_input_tokens": 128000, "max_output_tokens": 128000, - "input_cost_per_token": 0.000000125, - "output_cost_per_token": 0.00000045, + "input_cost_per_token": 1.25e-07, + "output_cost_per_token": 4.5e-07, "litellm_provider": "pinstripes", "mode": "chat", "supports_function_calling": true, @@ -43553,8 +43791,8 @@ "max_tokens": 131072, "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 0.00000014, - "output_cost_per_token": 0.00000045, + "input_cost_per_token": 1.4e-07, + "output_cost_per_token": 4.5e-07, "litellm_provider": "pinstripes", "mode": "chat", "supports_function_calling": true, @@ -43566,8 +43804,8 @@ "max_tokens": 131072, "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 0.00000009, - "output_cost_per_token": 0.0000002, + "input_cost_per_token": 9e-08, + "output_cost_per_token": 2e-07, "litellm_provider": "pinstripes", "mode": "chat", "supports_function_calling": true, @@ -43579,8 +43817,8 @@ "max_tokens": 131072, "max_input_tokens": 131072, "max_output_tokens": 131072, - "input_cost_per_token": 0.0000003, - "output_cost_per_token": 0.0000006, + "input_cost_per_token": 3e-07, + "output_cost_per_token": 6e-07, "litellm_provider": "pinstripes", "mode": "chat", "supports_function_calling": true, @@ -43592,8 +43830,8 @@ "max_tokens": 163840, "max_input_tokens": 163840, "max_output_tokens": 163840, - "input_cost_per_token": 0.0000001, - "output_cost_per_token": 0.0000002, + "input_cost_per_token": 1e-07, + "output_cost_per_token": 2e-07, "litellm_provider": "pinstripes", "mode": "chat", "supports_function_calling": true, @@ -43605,8 +43843,8 @@ "max_tokens": 1000192, "max_input_tokens": 1000192, "max_output_tokens": 1000192, - "input_cost_per_token": 0.000000255, - "output_cost_per_token": 0.00000055, + "input_cost_per_token": 2.55e-07, + "output_cost_per_token": 5.5e-07, "litellm_provider": "pinstripes", "mode": "chat", "supports_function_calling": true, @@ -43648,39 +43886,39 @@ "supports_system_messages": true, "supports_tool_choice": true }, - "fallback_generalizations": { - "rules": [ - { - "name": "anthropic-claude-adaptive-thinking", - "pattern": "(?:opus|sonnet|haiku)[-._](?:4[-._](?:[6-9]|[1-9]\\d)(?!\\d)|(?:[5-9]|[1-9]\\d{1,})[-._]\\d{1,2}(?!\\d))", - "description": "Claude opus/sonnet/haiku at version 4.6 or higher: 4.6 through 4.99, then any 5.x, 6.x or later major. The minor is capped at two digits so an 8-digit date suffix such as claude-opus-4-20250514 is never read as a >= 4.6 minor. Turns on adaptive thinking for new families with no code change.", - "extends": "anthropic-claude", - "model_info": { - "supports_adaptive_thinking": true - } - }, - { - "name": "anthropic-claude", - "pattern": "^claude-[a-z]+-\\d+[-.]\\d+(?:-\\d{8})?$", - "description": "Any Claude family-major-minor id, optionally with an 8-digit date suffix, anchored to the whole name. Version-neutral fallback that gives an unmapped Claude provider routing and baseline capabilities; it carries no pricing, so cost stays on the standard unpriced behavior rather than a guessed number.", - "model_info": { - "litellm_provider": "anthropic", - "mode": "chat", - "max_input_tokens": 200000, - "max_output_tokens": 64000, - "max_tokens": 64000, - "supports_function_calling": true, - "supports_parallel_function_calling": true, - "supports_vision": true, - "supports_tool_choice": true, - "supports_assistant_prefill": true, - "supports_prompt_caching": true, - "supports_response_schema": true, - "supports_reasoning": true, - "supports_pdf_input": true, - "supports_system_messages": true - } - } - ] - } + "fallback_generalizations": { + "rules": [ + { + "name": "anthropic-claude-adaptive-thinking", + "pattern": "(?:opus|sonnet|haiku)[-._](?:4[-._](?:[6-9]|[1-9]\\d)(?!\\d)|(?:[5-9]|[1-9]\\d{1,})[-._]\\d{1,2}(?!\\d))", + "description": "Claude opus/sonnet/haiku at version 4.6 or higher: 4.6 through 4.99, then any 5.x, 6.x or later major. The minor is capped at two digits so an 8-digit date suffix such as claude-opus-4-20250514 is never read as a >= 4.6 minor. Turns on adaptive thinking for new families with no code change.", + "extends": "anthropic-claude", + "model_info": { + "supports_adaptive_thinking": true + } + }, + { + "name": "anthropic-claude", + "pattern": "^claude-[a-z]+-\\d+[-.]\\d+(?:-\\d{8})?$", + "description": "Any Claude family-major-minor id, optionally with an 8-digit date suffix, anchored to the whole name. Version-neutral fallback that gives an unmapped Claude provider routing and baseline capabilities; it carries no pricing, so cost stays on the standard unpriced behavior rather than a guessed number.", + "model_info": { + "litellm_provider": "anthropic", + "mode": "chat", + "max_input_tokens": 200000, + "max_output_tokens": 64000, + "max_tokens": 64000, + "supports_function_calling": true, + "supports_parallel_function_calling": true, + "supports_vision": true, + "supports_tool_choice": true, + "supports_assistant_prefill": true, + "supports_prompt_caching": true, + "supports_response_schema": true, + "supports_reasoning": true, + "supports_pdf_input": true, + "supports_system_messages": true + } + } + ] + } } diff --git a/tests/test_litellm/test_claude_sonnet_5_config.py b/tests/test_litellm/test_claude_sonnet_5_config.py new file mode 100644 index 00000000000..943c4172205 --- /dev/null +++ b/tests/test_litellm/test_claude_sonnet_5_config.py @@ -0,0 +1,163 @@ +""" +Validate Claude Sonnet 5 model configuration entries. + +Sonnet 5 launched 2026-06-29 with introductory pricing ($2/$10 per MTok) +through August 31, 2026, after which it moves to standard pricing ($3/$15). +The cost-map entries use the introductory rates so spend tracking is accurate +during the promotional period. +""" + +import json +import os + +import pytest + +import litellm +from litellm.litellm_core_utils.get_model_cost_map import GetModelCostMap + +REPO_ROOT = os.path.join(os.path.dirname(__file__), "../..") + + +def _load_root_cost_map() -> dict: + json_path = os.path.join(REPO_ROOT, "model_prices_and_context_window.json") + with open(json_path) as f: + return json.load(f) + + +@pytest.fixture +def local_model_cost_map(monkeypatch): + original_model_cost = litellm.model_cost + monkeypatch.setenv("LITELLM_LOCAL_MODEL_COST_MAP", "True") + litellm.model_cost = litellm.get_model_cost_map(url="") + litellm.get_model_info.cache_clear() + try: + yield + finally: + litellm.model_cost = original_model_cost + litellm.get_model_info.cache_clear() + + +def test_sonnet_5_model_pricing_and_capabilities(): + model_data = _load_root_cost_map() + + expected_models = [ + ("claude-sonnet-5", "anthropic"), + ("anthropic.claude-sonnet-5", "bedrock_converse"), + ("vertex_ai/claude-sonnet-5", "vertex_ai-anthropic_models"), + ("azure_ai/claude-sonnet-5", "azure_ai"), + ] + + for model_name, provider in expected_models: + assert model_name in model_data, f"Missing model entry: {model_name}" + info = model_data[model_name] + + assert info["litellm_provider"] == provider + assert info["mode"] == "chat" + assert info["max_input_tokens"] == 1000000 + assert info["max_output_tokens"] == 128000 + assert info["max_tokens"] == 128000 + + assert info["input_cost_per_token"] == 2e-06 + assert info["output_cost_per_token"] == 1e-05 + assert info["cache_creation_input_token_cost"] == 2.5e-06 + assert info["cache_creation_input_token_cost_above_1hr"] == 4e-06 + assert info["cache_read_input_token_cost"] == 2e-07 + + assert "input_cost_per_token_above_200k_tokens" not in info + assert "output_cost_per_token_above_200k_tokens" not in info + + assert info["supports_adaptive_thinking"] is True + assert info["supports_function_calling"] is True + assert info["supports_prompt_caching"] is True + assert info["supports_reasoning"] is True + assert info["supports_tool_choice"] is True + assert info["supports_vision"] is True + + +def test_sonnet_5_bedrock_regional_model_pricing(): + model_data = _load_root_cost_map() + + expected_models = { + "global.anthropic.claude-sonnet-5": { + "input_cost_per_token": 2e-06, + "output_cost_per_token": 1e-05, + "cache_creation_input_token_cost": 2.5e-06, + "cache_read_input_token_cost": 2e-07, + }, + "us.anthropic.claude-sonnet-5": { + "input_cost_per_token": 2.2e-06, + "output_cost_per_token": 1.1e-05, + "cache_creation_input_token_cost": 2.75e-06, + "cache_read_input_token_cost": 2.2e-07, + }, + "eu.anthropic.claude-sonnet-5": { + "input_cost_per_token": 2.2e-06, + "output_cost_per_token": 1.1e-05, + "cache_creation_input_token_cost": 2.75e-06, + "cache_read_input_token_cost": 2.2e-07, + }, + "au.anthropic.claude-sonnet-5": { + "input_cost_per_token": 2.2e-06, + "output_cost_per_token": 1.1e-05, + "cache_creation_input_token_cost": 2.75e-06, + "cache_read_input_token_cost": 2.2e-07, + }, + "jp.anthropic.claude-sonnet-5": { + "input_cost_per_token": 2.2e-06, + "output_cost_per_token": 1.1e-05, + "cache_creation_input_token_cost": 2.75e-06, + "cache_read_input_token_cost": 2.2e-07, + }, + } + + for model_name, expected in expected_models.items(): + assert model_name in model_data, f"Missing model entry: {model_name}" + info = model_data[model_name] + assert info["litellm_provider"] == "bedrock_converse" + assert info["max_input_tokens"] == 1000000 + assert info["max_output_tokens"] == 128000 + for key, value in expected.items(): + assert info[key] == value, f"{model_name}.{key}: expected {value}, got {info[key]}" + + +def test_sonnet_5_present_in_bundled_backup(): + backup = GetModelCostMap.load_local_model_cost_map() + root = _load_root_cost_map() + for model_name in ( + "claude-sonnet-5", + "anthropic.claude-sonnet-5", + "global.anthropic.claude-sonnet-5", + "us.anthropic.claude-sonnet-5", + "eu.anthropic.claude-sonnet-5", + "au.anthropic.claude-sonnet-5", + "jp.anthropic.claude-sonnet-5", + "vertex_ai/claude-sonnet-5", + "vertex_ai/claude-sonnet-5@default", + "azure_ai/claude-sonnet-5", + "snowflake/claude-sonnet-5", + ): + assert model_name in backup, f"Missing from backup cost map: {model_name}" + assert backup[model_name] == root[model_name], model_name + + +def test_sonnet_5_provider_resolves_via_model_info(local_model_cost_map): + info = litellm.get_model_info(model="claude-sonnet-5") + assert info["litellm_provider"] == "anthropic" + assert info["max_input_tokens"] == 1000000 + assert info["max_output_tokens"] == 128000 + assert info["input_cost_per_token"] == 2e-06 + assert info["output_cost_per_token"] == 1e-05 + + +@pytest.mark.parametrize( + "cost_map", + [_load_root_cost_map(), GetModelCostMap.load_local_model_cost_map()], + ids=["root", "bundled_backup"], +) +def test_sonnet_5_all_variants_carry_adaptive_thinking_flag(cost_map): + variants = [k for k in cost_map if "claude-sonnet-5" in k and "sonnet-4-5" not in k] + assert variants, "no claude-sonnet-5 entries found in cost map" + missing = [ + k for k in variants if cost_map[k].get("supports_adaptive_thinking") is not True + ] + assert not missing, f"missing supports_adaptive_thinking: {missing}" From 901a097b86ddb9539982173bcc91a6a29a27a2d7 Mon Sep 17 00:00:00 2001 From: unknown <> Date: Wed, 1 Jul 2026 14:35:36 +0000 Subject: [PATCH 02/11] fix: restore supports_reasoning on minimax entries lost during dedup The original JSON had duplicate minimax.minimax-m2.5 entries; the first had supports_reasoning: true, the second did not. JSON deduplication kept the second. Restore the flag on the surviving entries and add 1hr cache TTL assertions to the regional bedrock test Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- litellm/model_prices_and_context_window_backup.json | 9 ++++++--- model_prices_and_context_window.json | 9 ++++++--- tests/test_litellm/test_claude_sonnet_5_config.py | 5 +++++ 3 files changed, 17 insertions(+), 6 deletions(-) diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index 15b16118a7b..dba8bb949d0 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -9325,7 +9325,8 @@ "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "source": "https://aws.amazon.com/bedrock/pricing/", + "supports_reasoning": true }, "bedrock/us-east-1/moonshotai.kimi-k2-thinking": { "input_cost_per_token": 6e-07, @@ -9946,7 +9947,8 @@ "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "source": "https://aws.amazon.com/bedrock/pricing/", + "supports_reasoning": true }, "bedrock/us-west-2/moonshotai.kimi-k2-thinking": { "input_cost_per_token": 6e-07, @@ -25092,7 +25094,8 @@ "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "source": "https://aws.amazon.com/bedrock/pricing/", + "supports_reasoning": true }, "minimax/speech-02-hd": { "input_cost_per_character": 0.0001, diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index 15b16118a7b..dba8bb949d0 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -9325,7 +9325,8 @@ "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "source": "https://aws.amazon.com/bedrock/pricing/", + "supports_reasoning": true }, "bedrock/us-east-1/moonshotai.kimi-k2-thinking": { "input_cost_per_token": 6e-07, @@ -9946,7 +9947,8 @@ "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "source": "https://aws.amazon.com/bedrock/pricing/", + "supports_reasoning": true }, "bedrock/us-west-2/moonshotai.kimi-k2-thinking": { "input_cost_per_token": 6e-07, @@ -25092,7 +25094,8 @@ "supports_function_calling": true, "supports_system_messages": true, "supports_tool_choice": true, - "source": "https://aws.amazon.com/bedrock/pricing/" + "source": "https://aws.amazon.com/bedrock/pricing/", + "supports_reasoning": true }, "minimax/speech-02-hd": { "input_cost_per_character": 0.0001, diff --git a/tests/test_litellm/test_claude_sonnet_5_config.py b/tests/test_litellm/test_claude_sonnet_5_config.py index 943c4172205..c8726a039f2 100644 --- a/tests/test_litellm/test_claude_sonnet_5_config.py +++ b/tests/test_litellm/test_claude_sonnet_5_config.py @@ -82,30 +82,35 @@ def test_sonnet_5_bedrock_regional_model_pricing(): "input_cost_per_token": 2e-06, "output_cost_per_token": 1e-05, "cache_creation_input_token_cost": 2.5e-06, + "cache_creation_input_token_cost_above_1hr": 4e-06, "cache_read_input_token_cost": 2e-07, }, "us.anthropic.claude-sonnet-5": { "input_cost_per_token": 2.2e-06, "output_cost_per_token": 1.1e-05, "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, }, "eu.anthropic.claude-sonnet-5": { "input_cost_per_token": 2.2e-06, "output_cost_per_token": 1.1e-05, "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, }, "au.anthropic.claude-sonnet-5": { "input_cost_per_token": 2.2e-06, "output_cost_per_token": 1.1e-05, "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, }, "jp.anthropic.claude-sonnet-5": { "input_cost_per_token": 2.2e-06, "output_cost_per_token": 1.1e-05, "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, }, } From bbf51146ca7ffa63e5a46c903fa13b14b7f48c59 Mon Sep 17 00:00:00 2001 From: mateo Date: Sat, 5 Sep 2026 13:29:38 +0000 Subject: [PATCH 03/11] fix(pricing): correct Vertex Haiku 4.5 output limit, Bedrock Mantle gpt-oss rates, add Scaleway deprecation dates Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- ...odel_prices_and_context_window_backup.json | 32 +++++++++++-------- model_prices_and_context_window.json | 32 +++++++++++-------- .../test_bedrock_mantle_transformation.py | 4 +-- 3 files changed, 40 insertions(+), 28 deletions(-) diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index 2459ed940e0..1bc4c275161 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -45421,8 +45421,8 @@ "input_cost_per_token": 1e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 200000, - "max_output_tokens": 8192, - "max_tokens": 8192, + "max_output_tokens": 64000, + "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, "regional_endpoint_uplift_multiplier": 1.1, @@ -45446,8 +45446,8 @@ "input_cost_per_token": 1e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 200000, - "max_output_tokens": 8192, - "max_tokens": 8192, + "max_output_tokens": 64000, + "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, "regional_endpoint_uplift_multiplier": 1.1, @@ -51727,7 +51727,8 @@ "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 8e-07, - "supports_function_calling": true + "supports_function_calling": true, + "deprecation_date": "2026-10-01" }, "scaleway/openai/gpt-oss-120b": { "input_cost_per_token": 1.5e-07, @@ -51766,7 +51767,8 @@ "mode": "chat", "output_cost_per_token": 5e-07, "supports_function_calling": true, - "supports_vision": true + "supports_vision": true, + "deprecation_date": "2026-08-01" }, "scaleway/hcompany/holo2-30b-a3b": { "input_cost_per_token": 3e-07, @@ -51777,7 +51779,8 @@ "mode": "chat", "output_cost_per_token": 7e-07, "supports_reasoning": true, - "supports_vision": true + "supports_vision": true, + "deprecation_date": "2026-08-09" }, "scaleway/mistralai/mistral-medium-3.5-128b": { "input_cost_per_token": 1.5e-06, @@ -51800,7 +51803,8 @@ "max_tokens": 16384, "mode": "chat", "output_cost_per_token": 2e-06, - "supports_function_calling": true + "supports_function_calling": true, + "deprecation_date": "2026-08-01" }, "scaleway/mistralai/voxtral-small-24b-2507": { "input_cost_per_audio_token": 1.5e-07, @@ -51811,7 +51815,8 @@ "max_tokens": 16384, "mode": "chat", "output_cost_per_token": 3.5e-07, - "supports_audio_input": true + "supports_audio_input": true, + "deprecation_date": "2026-08-01" }, "scaleway/mistralai/mistral-small-3.2-24b-instruct-2506": { "input_cost_per_token": 1.5e-07, @@ -51833,7 +51838,8 @@ "mode": "chat", "output_cost_per_token": 2e-07, "supports_vision": true, - "supports_function_calling": true + "supports_function_calling": true, + "deprecation_date": "2026-10-01" }, "scaleway/BAAI/bge-multilingual-gemma2": { "input_cost_per_token": 1e-07, @@ -54283,7 +54289,7 @@ "supports_tool_choice": true }, "bedrock_mantle/openai.gpt-oss-20b": { - "input_cost_per_token": 7.5e-08, + "input_cost_per_token": 7e-08, "output_cost_per_token": 3e-07, "litellm_provider": "bedrock_mantle", "max_input_tokens": 131072, @@ -54317,8 +54323,8 @@ "supports_tool_choice": true }, "bedrock_mantle/openai.gpt-oss-safeguard-20b": { - "input_cost_per_token": 7.5e-08, - "output_cost_per_token": 3e-07, + "input_cost_per_token": 7e-08, + "output_cost_per_token": 2e-07, "litellm_provider": "bedrock_mantle", "max_input_tokens": 131072, "max_output_tokens": 65536, diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index 2459ed940e0..1bc4c275161 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -45421,8 +45421,8 @@ "input_cost_per_token": 1e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 200000, - "max_output_tokens": 8192, - "max_tokens": 8192, + "max_output_tokens": 64000, + "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, "regional_endpoint_uplift_multiplier": 1.1, @@ -45446,8 +45446,8 @@ "input_cost_per_token": 1e-06, "litellm_provider": "vertex_ai-anthropic_models", "max_input_tokens": 200000, - "max_output_tokens": 8192, - "max_tokens": 8192, + "max_output_tokens": 64000, + "max_tokens": 64000, "mode": "chat", "output_cost_per_token": 5e-06, "regional_endpoint_uplift_multiplier": 1.1, @@ -51727,7 +51727,8 @@ "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 8e-07, - "supports_function_calling": true + "supports_function_calling": true, + "deprecation_date": "2026-10-01" }, "scaleway/openai/gpt-oss-120b": { "input_cost_per_token": 1.5e-07, @@ -51766,7 +51767,8 @@ "mode": "chat", "output_cost_per_token": 5e-07, "supports_function_calling": true, - "supports_vision": true + "supports_vision": true, + "deprecation_date": "2026-08-01" }, "scaleway/hcompany/holo2-30b-a3b": { "input_cost_per_token": 3e-07, @@ -51777,7 +51779,8 @@ "mode": "chat", "output_cost_per_token": 7e-07, "supports_reasoning": true, - "supports_vision": true + "supports_vision": true, + "deprecation_date": "2026-08-09" }, "scaleway/mistralai/mistral-medium-3.5-128b": { "input_cost_per_token": 1.5e-06, @@ -51800,7 +51803,8 @@ "max_tokens": 16384, "mode": "chat", "output_cost_per_token": 2e-06, - "supports_function_calling": true + "supports_function_calling": true, + "deprecation_date": "2026-08-01" }, "scaleway/mistralai/voxtral-small-24b-2507": { "input_cost_per_audio_token": 1.5e-07, @@ -51811,7 +51815,8 @@ "max_tokens": 16384, "mode": "chat", "output_cost_per_token": 3.5e-07, - "supports_audio_input": true + "supports_audio_input": true, + "deprecation_date": "2026-08-01" }, "scaleway/mistralai/mistral-small-3.2-24b-instruct-2506": { "input_cost_per_token": 1.5e-07, @@ -51833,7 +51838,8 @@ "mode": "chat", "output_cost_per_token": 2e-07, "supports_vision": true, - "supports_function_calling": true + "supports_function_calling": true, + "deprecation_date": "2026-10-01" }, "scaleway/BAAI/bge-multilingual-gemma2": { "input_cost_per_token": 1e-07, @@ -54283,7 +54289,7 @@ "supports_tool_choice": true }, "bedrock_mantle/openai.gpt-oss-20b": { - "input_cost_per_token": 7.5e-08, + "input_cost_per_token": 7e-08, "output_cost_per_token": 3e-07, "litellm_provider": "bedrock_mantle", "max_input_tokens": 131072, @@ -54317,8 +54323,8 @@ "supports_tool_choice": true }, "bedrock_mantle/openai.gpt-oss-safeguard-20b": { - "input_cost_per_token": 7.5e-08, - "output_cost_per_token": 3e-07, + "input_cost_per_token": 7e-08, + "output_cost_per_token": 2e-07, "litellm_provider": "bedrock_mantle", "max_input_tokens": 131072, "max_output_tokens": 65536, diff --git a/tests/test_litellm/llms/bedrock_mantle/test_bedrock_mantle_transformation.py b/tests/test_litellm/llms/bedrock_mantle/test_bedrock_mantle_transformation.py index b88c27e64b9..92c012d339c 100644 --- a/tests/test_litellm/llms/bedrock_mantle/test_bedrock_mantle_transformation.py +++ b/tests/test_litellm/llms/bedrock_mantle/test_bedrock_mantle_transformation.py @@ -696,8 +696,8 @@ class TestBedrockMantlePricing: monkeypatch.setenv("LITELLM_LOCAL_MODEL_COST_MAP", "true") litellm.add_known_models() info = litellm.get_model_info("bedrock_mantle/openai.gpt-oss-20b") - # Bedrock pricing: $0.075/M input, $0.30/M output - assert info["input_cost_per_token"] == pytest.approx(7.5e-8) + # Bedrock pricing: $0.07/M input, $0.30/M output + assert info["input_cost_per_token"] == pytest.approx(7e-8) assert info["output_cost_per_token"] == pytest.approx(3e-7) def test_pricing_significantly_cheaper_than_openai_native(self, monkeypatch): From 528f3ae0161192004ca76d615363c748dcfdcd0d Mon Sep 17 00:00:00 2001 From: mateo Date: Sat, 5 Sep 2026 13:46:46 +0000 Subject: [PATCH 04/11] test(bedrock_mantle): cover gpt-oss-safeguard-20b input and output rates Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- .../bedrock_mantle/test_bedrock_mantle_transformation.py | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/tests/test_litellm/llms/bedrock_mantle/test_bedrock_mantle_transformation.py b/tests/test_litellm/llms/bedrock_mantle/test_bedrock_mantle_transformation.py index 92c012d339c..b3eec97058a 100644 --- a/tests/test_litellm/llms/bedrock_mantle/test_bedrock_mantle_transformation.py +++ b/tests/test_litellm/llms/bedrock_mantle/test_bedrock_mantle_transformation.py @@ -700,6 +700,14 @@ class TestBedrockMantlePricing: assert info["input_cost_per_token"] == pytest.approx(7e-8) assert info["output_cost_per_token"] == pytest.approx(3e-7) + def test_gpt_oss_safeguard_20b_pricing(self, monkeypatch): + monkeypatch.setenv("LITELLM_LOCAL_MODEL_COST_MAP", "true") + litellm.add_known_models() + info = litellm.get_model_info("bedrock_mantle/openai.gpt-oss-safeguard-20b") + # Bedrock pricing: $0.07/M input, $0.20/M output + assert info["input_cost_per_token"] == pytest.approx(7e-8) + assert info["output_cost_per_token"] == pytest.approx(2e-7) + def test_pricing_significantly_cheaper_than_openai_native(self, monkeypatch): """ Verify Bedrock Mantle pricing is cheaper than OpenAI's direct API pricing. From b4e7776ab3534b4c6b91073aea1a1bbf8f681e83 Mon Sep 17 00:00:00 2001 From: mateo Date: Sat, 5 Sep 2026 19:21:18 +0000 Subject: [PATCH 05/11] fix(model_prices): add gemini lyria-3.5, voyage-multilingual-2, chatgpt gpt-5.5 and gpt-5.6 entries Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- ...odel_prices_and_context_window_backup.json | 99 +++++++++++++++++++ model_prices_and_context_window.json | 99 +++++++++++++++++++ .../test_chatgpt_responses_transformation.py | 53 +++++++++- 3 files changed, 249 insertions(+), 2 deletions(-) diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index 1bc4c275161..8780c41b70a 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -27541,6 +27541,70 @@ "max_tokens": 8191, "mode": "embedding" }, + "chatgpt/gpt-5.5": { + "litellm_provider": "chatgpt", + "source": "https://platform.openai.com/docs/models/gpt-5.5", + "max_input_tokens": 1050000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "responses", + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], + "supports_function_calling": true, + "supports_parallel_function_calling": true, + "supports_response_schema": true, + "supports_vision": true + }, + "chatgpt/gpt-5.6-luna": { + "litellm_provider": "chatgpt", + "source": "https://platform.openai.com/docs/models/gpt-5.6-luna", + "max_input_tokens": 1050000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "responses", + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], + "supports_function_calling": true, + "supports_parallel_function_calling": true, + "supports_response_schema": true, + "supports_vision": true + }, + "chatgpt/gpt-5.6-sol": { + "litellm_provider": "chatgpt", + "source": "https://platform.openai.com/docs/models/gpt-5.6-sol", + "max_input_tokens": 1050000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "responses", + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], + "supports_function_calling": true, + "supports_parallel_function_calling": true, + "supports_response_schema": true, + "supports_vision": true + }, + "chatgpt/gpt-5.6-terra": { + "litellm_provider": "chatgpt", + "source": "https://platform.openai.com/docs/models/gpt-5.6-terra", + "max_input_tokens": 1050000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "responses", + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], + "supports_function_calling": true, + "supports_parallel_function_calling": true, + "supports_response_schema": true, + "supports_vision": true + }, "chatgpt/gpt-5.4": { "litellm_provider": "chatgpt", "max_input_tokens": 1050000, @@ -47654,6 +47718,16 @@ "mode": "embedding", "output_cost_per_token": 0.0 }, + "voyage/voyage-multilingual-2": { + "input_cost_per_token": 1.2e-07, + "litellm_provider": "voyage", + "max_input_tokens": 32000, + "max_tokens": 32000, + "mode": "embedding", + "output_cost_per_token": 0.0, + "output_vector_size": 1024, + "source": "https://docs.voyageai.com/docs/pricing" + }, "voyage/voyage-3-large": { "input_cost_per_token": 1.8e-07, "litellm_provider": "voyage", @@ -60044,6 +60118,31 @@ "supports_web_search": false, "output_cost_per_image": 0.08 }, + "gemini/lyria-3.5": { + "input_cost_per_token": 0, + "litellm_provider": "gemini", + "max_input_tokens": 1048576, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 0, + "source": "https://ai.google.dev/gemini-api/docs/pricing", + "supported_modalities": [ + "text" + ], + "supported_output_modalities": [ + "audio" + ], + "supports_audio_input": false, + "supports_audio_output": true, + "supports_function_calling": false, + "supports_prompt_caching": false, + "supports_response_schema": false, + "supports_system_messages": false, + "supports_vision": false, + "supports_web_search": false, + "output_cost_per_image": 0.08 + }, "perplexity/anthropic/claude-fable-5": { "litellm_provider": "perplexity", "mode": "responses", diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index 1bc4c275161..8780c41b70a 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -27541,6 +27541,70 @@ "max_tokens": 8191, "mode": "embedding" }, + "chatgpt/gpt-5.5": { + "litellm_provider": "chatgpt", + "source": "https://platform.openai.com/docs/models/gpt-5.5", + "max_input_tokens": 1050000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "responses", + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], + "supports_function_calling": true, + "supports_parallel_function_calling": true, + "supports_response_schema": true, + "supports_vision": true + }, + "chatgpt/gpt-5.6-luna": { + "litellm_provider": "chatgpt", + "source": "https://platform.openai.com/docs/models/gpt-5.6-luna", + "max_input_tokens": 1050000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "responses", + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], + "supports_function_calling": true, + "supports_parallel_function_calling": true, + "supports_response_schema": true, + "supports_vision": true + }, + "chatgpt/gpt-5.6-sol": { + "litellm_provider": "chatgpt", + "source": "https://platform.openai.com/docs/models/gpt-5.6-sol", + "max_input_tokens": 1050000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "responses", + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], + "supports_function_calling": true, + "supports_parallel_function_calling": true, + "supports_response_schema": true, + "supports_vision": true + }, + "chatgpt/gpt-5.6-terra": { + "litellm_provider": "chatgpt", + "source": "https://platform.openai.com/docs/models/gpt-5.6-terra", + "max_input_tokens": 1050000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "responses", + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], + "supports_function_calling": true, + "supports_parallel_function_calling": true, + "supports_response_schema": true, + "supports_vision": true + }, "chatgpt/gpt-5.4": { "litellm_provider": "chatgpt", "max_input_tokens": 1050000, @@ -47654,6 +47718,16 @@ "mode": "embedding", "output_cost_per_token": 0.0 }, + "voyage/voyage-multilingual-2": { + "input_cost_per_token": 1.2e-07, + "litellm_provider": "voyage", + "max_input_tokens": 32000, + "max_tokens": 32000, + "mode": "embedding", + "output_cost_per_token": 0.0, + "output_vector_size": 1024, + "source": "https://docs.voyageai.com/docs/pricing" + }, "voyage/voyage-3-large": { "input_cost_per_token": 1.8e-07, "litellm_provider": "voyage", @@ -60044,6 +60118,31 @@ "supports_web_search": false, "output_cost_per_image": 0.08 }, + "gemini/lyria-3.5": { + "input_cost_per_token": 0, + "litellm_provider": "gemini", + "max_input_tokens": 1048576, + "max_output_tokens": 65536, + "max_tokens": 65536, + "mode": "chat", + "output_cost_per_token": 0, + "source": "https://ai.google.dev/gemini-api/docs/pricing", + "supported_modalities": [ + "text" + ], + "supported_output_modalities": [ + "audio" + ], + "supports_audio_input": false, + "supports_audio_output": true, + "supports_function_calling": false, + "supports_prompt_caching": false, + "supports_response_schema": false, + "supports_system_messages": false, + "supports_vision": false, + "supports_web_search": false, + "output_cost_per_image": 0.08 + }, "perplexity/anthropic/claude-fable-5": { "litellm_provider": "perplexity", "mode": "responses", diff --git a/tests/test_litellm/llms/chatgpt/responses/test_chatgpt_responses_transformation.py b/tests/test_litellm/llms/chatgpt/responses/test_chatgpt_responses_transformation.py index 8e0415d50de..8a0027fdb53 100644 --- a/tests/test_litellm/llms/chatgpt/responses/test_chatgpt_responses_transformation.py +++ b/tests/test_litellm/llms/chatgpt/responses/test_chatgpt_responses_transformation.py @@ -10,18 +10,23 @@ from unittest.mock import MagicMock, patch import httpx import pytest - +import litellm +from litellm.llms.chatgpt.responses.transformation import ChatGPTResponsesAPIConfig from litellm.llms.openai.common_utils import OpenAIError +from litellm.main import responses_api_bridge_check from litellm.types.router import GenericLiteLLMParams from litellm.types.utils import LlmProviders from litellm.utils import ProviderConfigManager -from litellm.llms.chatgpt.responses.transformation import ChatGPTResponsesAPIConfig class TestChatGPTResponsesAPITransformation: @pytest.mark.parametrize( "model_name", [ + "chatgpt/gpt-5.5", + "chatgpt/gpt-5.6-luna", + "chatgpt/gpt-5.6-sol", + "chatgpt/gpt-5.6-terra", "chatgpt/gpt-5.4", "chatgpt/gpt-5.4-pro", "chatgpt/gpt-5.3-chat-latest", @@ -40,6 +45,50 @@ class TestChatGPTResponsesAPITransformation: assert isinstance(config, ChatGPTResponsesAPIConfig) assert config.custom_llm_provider == LlmProviders.CHATGPT + @pytest.mark.parametrize( + "model_name", + [ + "chatgpt/gpt-5.5", + "chatgpt/gpt-5.6-luna", + "chatgpt/gpt-5.6-sol", + "chatgpt/gpt-5.6-terra", + ], + ) + def test_chatgpt_responses_model_metadata(self, model_name, local_model_cost_map): + model_info = litellm.get_model_info(model_name) + + assert model_info["litellm_provider"] == "chatgpt" + assert model_info["mode"] == "responses" + assert model_info["supported_endpoints"] == [ + "/v1/chat/completions", + "/v1/responses", + ] + assert model_info["max_input_tokens"] == 1050000 + assert model_info["max_output_tokens"] == 128000 + + @pytest.mark.parametrize( + "model_name", + [ + "gpt-5.5", + "gpt-5.6-luna", + "gpt-5.6-sol", + "gpt-5.6-terra", + ], + ) + def test_chatgpt_models_bridge_chat_completions_to_responses(self, model_name, local_model_cost_map): + """A chat completions request for these models must take the Responses bridge. + + `gpt-5.6-*` also exists as an openai chat model, so an unregistered + chatgpt model resolves to mode "chat" here and never reaches the bridge. + """ + model_info, resolved_model = responses_api_bridge_check( + model=model_name, + custom_llm_provider="chatgpt", + ) + + assert model_info["mode"] == "responses" + assert resolved_model == model_name + @patch("litellm.llms.chatgpt.responses.transformation.Authenticator") def test_chatgpt_responses_endpoint_url(self, mock_authenticator_class): mock_auth_instance = MagicMock() From 15c9fca53fc04c5ab05ebba0a8f047ef77394abd Mon Sep 17 00:00:00 2001 From: mateo Date: Mon, 7 Sep 2026 13:17:30 +0000 Subject: [PATCH 06/11] fix(pricing): add xai grok-imagine-video entries, computer-use-preview deprecation, gemini live 2.5 native audio card values Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- ...odel_prices_and_context_window_backup.json | 94 +++++++++++++++++-- model_prices_and_context_window.json | 94 +++++++++++++++++-- 2 files changed, 174 insertions(+), 14 deletions(-) diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index c558a953032..33e9ce94455 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -14336,6 +14336,7 @@ "supports_tool_choice": true }, "computer-use-preview": { + "deprecation_date": "2026-07-23", "input_cost_per_token": 3e-06, "litellm_provider": "azure", "max_input_tokens": 8192, @@ -23829,9 +23830,9 @@ "input_cost_per_audio_token": 3e-06, "input_cost_per_token": 5e-07, "litellm_provider": "vertex_ai-language-models", - "max_input_tokens": 1048576, - "max_output_tokens": 65535, - "max_tokens": 65535, + "max_input_tokens": 131072, + "max_output_tokens": 65536, + "max_tokens": 65536, "mode": "realtime", "output_cost_per_audio_token": 1.2e-05, "output_cost_per_token": 2e-06, @@ -23854,12 +23855,12 @@ "supports_audio_output": true, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_pdf_input": true, - "supports_prompt_caching": true, - "supports_response_schema": true, + "supports_pdf_input": false, + "supports_prompt_caching": false, + "supports_response_schema": false, "supports_system_messages": true, "supports_tool_choice": true, - "supports_url_context": true, + "supports_url_context": false, "supports_vision": true, "supports_web_search": true, "search_context_cost_per_query": { @@ -59479,6 +59480,85 @@ "image" ] }, + "xai/grok-imagine-video": { + "input_cost_per_image": 0.002, + "litellm_provider": "xai", + "mode": "video_generation", + "output_cost_per_second": 0.05, + "output_cost_per_second_480p": 0.05, + "source": "https://docs.x.ai/docs/models/grok-imagine-video", + "supported_endpoints": [ + "/v1/videos" + ], + "supported_modalities": [ + "text", + "image", + "video" + ], + "supported_output_modalities": [ + "video" + ] + }, + "xai/grok-imagine-video-1.5": { + "input_cost_per_image": 0.01, + "litellm_provider": "xai", + "mode": "video_generation", + "output_cost_per_second": 0.08, + "output_cost_per_second_1080p": 0.25, + "output_cost_per_second_480p": 0.08, + "source": "https://docs.x.ai/docs/models/grok-imagine-video-1.5", + "supported_endpoints": [ + "/v1/videos" + ], + "supported_modalities": [ + "text", + "image", + "audio" + ], + "supported_output_modalities": [ + "video" + ] + }, + "xai/grok-imagine-video-1.5-2026-05-30": { + "input_cost_per_image": 0.01, + "litellm_provider": "xai", + "mode": "video_generation", + "output_cost_per_second": 0.08, + "output_cost_per_second_1080p": 0.25, + "output_cost_per_second_480p": 0.08, + "source": "https://docs.x.ai/docs/models/grok-imagine-video-1.5", + "supported_endpoints": [ + "/v1/videos" + ], + "supported_modalities": [ + "text", + "image", + "audio" + ], + "supported_output_modalities": [ + "video" + ] + }, + "xai/grok-imagine-video-1.5-preview": { + "input_cost_per_image": 0.01, + "litellm_provider": "xai", + "mode": "video_generation", + "output_cost_per_second": 0.08, + "output_cost_per_second_1080p": 0.25, + "output_cost_per_second_480p": 0.08, + "source": "https://docs.x.ai/docs/models/grok-imagine-video-1.5", + "supported_endpoints": [ + "/v1/videos" + ], + "supported_modalities": [ + "text", + "image", + "audio" + ], + "supported_output_modalities": [ + "video" + ] + }, "low/1024-x-1024/grok-imagine-image-2.0": { "input_cost_per_image": 0.04, "litellm_provider": "xai", diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index c558a953032..33e9ce94455 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -14336,6 +14336,7 @@ "supports_tool_choice": true }, "computer-use-preview": { + "deprecation_date": "2026-07-23", "input_cost_per_token": 3e-06, "litellm_provider": "azure", "max_input_tokens": 8192, @@ -23829,9 +23830,9 @@ "input_cost_per_audio_token": 3e-06, "input_cost_per_token": 5e-07, "litellm_provider": "vertex_ai-language-models", - "max_input_tokens": 1048576, - "max_output_tokens": 65535, - "max_tokens": 65535, + "max_input_tokens": 131072, + "max_output_tokens": 65536, + "max_tokens": 65536, "mode": "realtime", "output_cost_per_audio_token": 1.2e-05, "output_cost_per_token": 2e-06, @@ -23854,12 +23855,12 @@ "supports_audio_output": true, "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_pdf_input": true, - "supports_prompt_caching": true, - "supports_response_schema": true, + "supports_pdf_input": false, + "supports_prompt_caching": false, + "supports_response_schema": false, "supports_system_messages": true, "supports_tool_choice": true, - "supports_url_context": true, + "supports_url_context": false, "supports_vision": true, "supports_web_search": true, "search_context_cost_per_query": { @@ -59479,6 +59480,85 @@ "image" ] }, + "xai/grok-imagine-video": { + "input_cost_per_image": 0.002, + "litellm_provider": "xai", + "mode": "video_generation", + "output_cost_per_second": 0.05, + "output_cost_per_second_480p": 0.05, + "source": "https://docs.x.ai/docs/models/grok-imagine-video", + "supported_endpoints": [ + "/v1/videos" + ], + "supported_modalities": [ + "text", + "image", + "video" + ], + "supported_output_modalities": [ + "video" + ] + }, + "xai/grok-imagine-video-1.5": { + "input_cost_per_image": 0.01, + "litellm_provider": "xai", + "mode": "video_generation", + "output_cost_per_second": 0.08, + "output_cost_per_second_1080p": 0.25, + "output_cost_per_second_480p": 0.08, + "source": "https://docs.x.ai/docs/models/grok-imagine-video-1.5", + "supported_endpoints": [ + "/v1/videos" + ], + "supported_modalities": [ + "text", + "image", + "audio" + ], + "supported_output_modalities": [ + "video" + ] + }, + "xai/grok-imagine-video-1.5-2026-05-30": { + "input_cost_per_image": 0.01, + "litellm_provider": "xai", + "mode": "video_generation", + "output_cost_per_second": 0.08, + "output_cost_per_second_1080p": 0.25, + "output_cost_per_second_480p": 0.08, + "source": "https://docs.x.ai/docs/models/grok-imagine-video-1.5", + "supported_endpoints": [ + "/v1/videos" + ], + "supported_modalities": [ + "text", + "image", + "audio" + ], + "supported_output_modalities": [ + "video" + ] + }, + "xai/grok-imagine-video-1.5-preview": { + "input_cost_per_image": 0.01, + "litellm_provider": "xai", + "mode": "video_generation", + "output_cost_per_second": 0.08, + "output_cost_per_second_1080p": 0.25, + "output_cost_per_second_480p": 0.08, + "source": "https://docs.x.ai/docs/models/grok-imagine-video-1.5", + "supported_endpoints": [ + "/v1/videos" + ], + "supported_modalities": [ + "text", + "image", + "audio" + ], + "supported_output_modalities": [ + "video" + ] + }, "low/1024-x-1024/grok-imagine-image-2.0": { "input_cost_per_image": 0.04, "litellm_provider": "xai", From 68d7a2d98fadd8f15eec9d5b0005e79de33e98bb Mon Sep 17 00:00:00 2001 From: mateo Date: Mon, 7 Sep 2026 13:29:29 +0000 Subject: [PATCH 07/11] fix(pricing): drop supported_endpoints from xai grok-imagine-video entries to match video model convention Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- litellm/model_prices_and_context_window_backup.json | 12 ------------ model_prices_and_context_window.json | 12 ------------ 2 files changed, 24 deletions(-) diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index 33e9ce94455..537db678f60 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -59487,9 +59487,6 @@ "output_cost_per_second": 0.05, "output_cost_per_second_480p": 0.05, "source": "https://docs.x.ai/docs/models/grok-imagine-video", - "supported_endpoints": [ - "/v1/videos" - ], "supported_modalities": [ "text", "image", @@ -59507,9 +59504,6 @@ "output_cost_per_second_1080p": 0.25, "output_cost_per_second_480p": 0.08, "source": "https://docs.x.ai/docs/models/grok-imagine-video-1.5", - "supported_endpoints": [ - "/v1/videos" - ], "supported_modalities": [ "text", "image", @@ -59527,9 +59521,6 @@ "output_cost_per_second_1080p": 0.25, "output_cost_per_second_480p": 0.08, "source": "https://docs.x.ai/docs/models/grok-imagine-video-1.5", - "supported_endpoints": [ - "/v1/videos" - ], "supported_modalities": [ "text", "image", @@ -59547,9 +59538,6 @@ "output_cost_per_second_1080p": 0.25, "output_cost_per_second_480p": 0.08, "source": "https://docs.x.ai/docs/models/grok-imagine-video-1.5", - "supported_endpoints": [ - "/v1/videos" - ], "supported_modalities": [ "text", "image", diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index 33e9ce94455..537db678f60 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -59487,9 +59487,6 @@ "output_cost_per_second": 0.05, "output_cost_per_second_480p": 0.05, "source": "https://docs.x.ai/docs/models/grok-imagine-video", - "supported_endpoints": [ - "/v1/videos" - ], "supported_modalities": [ "text", "image", @@ -59507,9 +59504,6 @@ "output_cost_per_second_1080p": 0.25, "output_cost_per_second_480p": 0.08, "source": "https://docs.x.ai/docs/models/grok-imagine-video-1.5", - "supported_endpoints": [ - "/v1/videos" - ], "supported_modalities": [ "text", "image", @@ -59527,9 +59521,6 @@ "output_cost_per_second_1080p": 0.25, "output_cost_per_second_480p": 0.08, "source": "https://docs.x.ai/docs/models/grok-imagine-video-1.5", - "supported_endpoints": [ - "/v1/videos" - ], "supported_modalities": [ "text", "image", @@ -59547,9 +59538,6 @@ "output_cost_per_second_1080p": 0.25, "output_cost_per_second_480p": 0.08, "source": "https://docs.x.ai/docs/models/grok-imagine-video-1.5", - "supported_endpoints": [ - "/v1/videos" - ], "supported_modalities": [ "text", "image", From cb8daee55f05e3d4ec728c1b3f15e153176721ea Mon Sep 17 00:00:00 2001 From: mateo Date: Tue, 8 Sep 2026 13:23:25 +0000 Subject: [PATCH 08/11] chore(registry): absorb #40185 daybreak-blue and #40159 cloudflare rpm limits Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- ...odel_prices_and_context_window_backup.json | 93 ++++++++++++++++--- model_prices_and_context_window.json | 93 ++++++++++++++++--- tests/test_litellm/test_utils.py | 6 ++ 3 files changed, 164 insertions(+), 28 deletions(-) diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index 5de376218b9..ef8da798922 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -13792,7 +13792,8 @@ "max_output_tokens": 3072, "max_tokens": 3072, "mode": "chat", - "output_cost_per_token": 1.923e-06 + "output_cost_per_token": 1.923e-06, + "rpm": 300 }, "cloudflare/@cf/meta/llama-2-7b-chat-int8": { "input_cost_per_token": 1.923e-06, @@ -13801,7 +13802,8 @@ "max_output_tokens": 2048, "max_tokens": 2048, "mode": "chat", - "output_cost_per_token": 1.923e-06 + "output_cost_per_token": 1.923e-06, + "rpm": 300 }, "cloudflare/@cf/mistral/mistral-7b-instruct-v0.1": { "input_cost_per_token": 1.923e-06, @@ -13810,7 +13812,8 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", - "output_cost_per_token": 1.923e-06 + "output_cost_per_token": 1.923e-06, + "rpm": 300 }, "cloudflare/@hf/thebloke/codellama-7b-instruct-awq": { "input_cost_per_token": 1.923e-06, @@ -13819,7 +13822,8 @@ "max_output_tokens": 4096, "max_tokens": 4096, "mode": "chat", - "output_cost_per_token": 1.923e-06 + "output_cost_per_token": 1.923e-06, + "rpm": 300 }, "cloudflare/@cf/openai/gpt-oss-120b": { "input_cost_per_token": 3.5e-07, @@ -13829,6 +13833,7 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 7.5e-07, + "rpm": 300, "supports_function_calling": true, "supports_reasoning": true }, @@ -13839,7 +13844,8 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", - "output_cost_per_token": 0.0 + "output_cost_per_token": 0.0, + "rpm": 300 }, "cloudflare/@cf/meta/llama-3.2-3b-instruct": { "input_cost_per_token": 5.09e-08, @@ -13848,7 +13854,8 @@ "max_output_tokens": 80000, "max_tokens": 80000, "mode": "chat", - "output_cost_per_token": 3.35e-07 + "output_cost_per_token": 3.35e-07, + "rpm": 300 }, "cloudflare/@cf/meta/llama-guard-3-8b": { "input_cost_per_token": 4.84e-07, @@ -13857,7 +13864,8 @@ "max_output_tokens": 131072, "max_tokens": 131072, "mode": "chat", - "output_cost_per_token": 3e-08 + "output_cost_per_token": 3e-08, + "rpm": 300 }, "cloudflare/@cf/mistral/mistral-7b-instruct-v0.2-lora": { "input_cost_per_token": 0.0, @@ -13866,7 +13874,8 @@ "max_output_tokens": 15000, "max_tokens": 15000, "mode": "chat", - "output_cost_per_token": 0.0 + "output_cost_per_token": 0.0, + "rpm": 300 }, "cloudflare/@cf/moonshotai/kimi-k2.7-code": { "cache_read_input_token_cost": 1.9e-07, @@ -13877,6 +13886,7 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 4e-06, + "rpm": 20, "supports_function_calling": true, "supports_reasoning": true }, @@ -13888,6 +13898,7 @@ "max_tokens": 80000, "mode": "chat", "output_cost_per_token": 4.881e-06, + "rpm": 300, "supports_reasoning": true }, "cloudflare/@cf/meta/llama-3.1-8b-instruct-fp8": { @@ -13897,7 +13908,8 @@ "max_output_tokens": 32000, "max_tokens": 32000, "mode": "chat", - "output_cost_per_token": 2.87e-07 + "output_cost_per_token": 2.87e-07, + "rpm": 300 }, "cloudflare/@cf/meta/llama-3.2-1b-instruct": { "input_cost_per_token": 2.7e-08, @@ -13906,7 +13918,8 @@ "max_output_tokens": 60000, "max_tokens": 60000, "mode": "chat", - "output_cost_per_token": 2.01e-07 + "output_cost_per_token": 2.01e-07, + "rpm": 300 }, "cloudflare/@cf/moonshotai/kimi-k2.6": { "cache_read_input_token_cost": 1.6e-07, @@ -13917,6 +13930,7 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 4e-06, + "rpm": 20, "supports_function_calling": true, "supports_reasoning": true }, @@ -13928,6 +13942,7 @@ "max_tokens": 131072, "mode": "chat", "output_cost_per_token": 4e-07, + "rpm": 300, "supports_function_calling": true, "supports_reasoning": true }, @@ -13938,7 +13953,8 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", - "output_cost_per_token": 0.0 + "output_cost_per_token": 0.0, + "rpm": 300 }, "cloudflare/@cf/meta/llama-3.3-70b-instruct-fp8-fast": { "input_cost_per_token": 2.93e-07, @@ -13948,6 +13964,7 @@ "max_tokens": 24000, "mode": "chat", "output_cost_per_token": 2.253e-06, + "rpm": 300, "supports_function_calling": true }, "cloudflare/@cf/ibm-granite/granite-4.0-h-micro": { @@ -13958,6 +13975,7 @@ "max_tokens": 131000, "mode": "chat", "output_cost_per_token": 1.12e-07, + "rpm": 300, "supports_function_calling": true }, "cloudflare/@cf/qwen/qwen2.5-coder-32b-instruct": { @@ -13967,7 +13985,8 @@ "max_output_tokens": 32768, "max_tokens": 32768, "mode": "chat", - "output_cost_per_token": 1e-06 + "output_cost_per_token": 1e-06, + "rpm": 300 }, "cloudflare/@cf/zai-org/glm-5.2": { "cache_read_input_token_cost": 2.6e-07, @@ -13978,6 +13997,7 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 4.4e-06, + "rpm": 20, "supports_function_calling": true, "supports_reasoning": true }, @@ -13989,6 +14009,7 @@ "max_tokens": 256000, "mode": "chat", "output_cost_per_token": 1.5e-06, + "rpm": 300, "supports_function_calling": true, "supports_reasoning": true }, @@ -13999,7 +14020,8 @@ "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", - "output_cost_per_token": 5.55e-07 + "output_cost_per_token": 5.55e-07, + "rpm": 300 }, "cloudflare/@cf/qwen/qwen3-30b-a3b-fp8": { "input_cost_per_token": 5.09e-08, @@ -14009,6 +14031,7 @@ "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 3.35e-07, + "rpm": 300, "supports_function_calling": true, "supports_reasoning": true }, @@ -14019,7 +14042,8 @@ "max_output_tokens": 3500, "max_tokens": 3500, "mode": "chat", - "output_cost_per_token": 0.0 + "output_cost_per_token": 0.0, + "rpm": 300 }, "cloudflare/@cf/google/gemma-4-26b-a4b-it": { "input_cost_per_token": 1e-07, @@ -14029,6 +14053,7 @@ "max_tokens": 256000, "mode": "chat", "output_cost_per_token": 3e-07, + "rpm": 300, "supports_function_calling": true, "supports_reasoning": true }, @@ -14040,6 +14065,7 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 5.55e-07, + "rpm": 300, "supports_function_calling": true }, "cloudflare/@cf/meta/llama-3.2-11b-vision-instruct": { @@ -14050,6 +14076,7 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 6.76e-07, + "rpm": 300, "supports_vision": true }, "cloudflare/@cf/openai/gpt-oss-20b": { @@ -14060,6 +14087,7 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 3e-07, + "rpm": 300, "supports_function_calling": true, "supports_reasoning": true }, @@ -14071,6 +14099,7 @@ "max_tokens": 131000, "mode": "chat", "output_cost_per_token": 8.5e-07, + "rpm": 300, "supports_function_calling": true }, "cloudflare/@cf/qwen/qwq-32b": { @@ -14081,6 +14110,7 @@ "max_tokens": 24000, "mode": "chat", "output_cost_per_token": 1e-06, + "rpm": 300, "supports_reasoning": true }, "codestral/codestral-2405": { @@ -54901,6 +54931,39 @@ "supports_tool_choice": true, "supports_vision": true }, + "bedrock_mantle/openai.gpt-daybreak-blue-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, + "litellm_provider": "bedrock_mantle", + "max_input_tokens": 1050000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "responses", + "use_openai_responses_path": true, + "supported_endpoints": [ + "/v1/responses" + ], + "supported_modalities": [ + "text", + "image" + ], + "supported_output_modalities": [ + "text" + ], + "supports_function_calling": true, + "supports_prompt_caching": 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-openai-gpt-daybreak-blue-56-sol.html" + }, "bedrock_mantle/openai.gpt-5.6-luna": { "input_cost_per_token": 2.2e-07, "input_cost_per_token_above_272k_tokens": 4.4e-07, @@ -60822,6 +60885,7 @@ "litellm_provider": "cloudflare", "mode": "audio_transcription", "output_cost_per_second": 0.0, + "rpm": 720, "source": "https://developers.cloudflare.com/workers-ai/models/whisper/", "supported_endpoints": [ "/v1/audio/transcriptions" @@ -60832,6 +60896,7 @@ "litellm_provider": "cloudflare", "mode": "audio_transcription", "output_cost_per_second": 0.0, + "rpm": 720, "source": "https://developers.cloudflare.com/workers-ai/models/whisper-large-v3-turbo/", "supported_endpoints": [ "/v1/audio/transcriptions" diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index 5de376218b9..ef8da798922 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -13792,7 +13792,8 @@ "max_output_tokens": 3072, "max_tokens": 3072, "mode": "chat", - "output_cost_per_token": 1.923e-06 + "output_cost_per_token": 1.923e-06, + "rpm": 300 }, "cloudflare/@cf/meta/llama-2-7b-chat-int8": { "input_cost_per_token": 1.923e-06, @@ -13801,7 +13802,8 @@ "max_output_tokens": 2048, "max_tokens": 2048, "mode": "chat", - "output_cost_per_token": 1.923e-06 + "output_cost_per_token": 1.923e-06, + "rpm": 300 }, "cloudflare/@cf/mistral/mistral-7b-instruct-v0.1": { "input_cost_per_token": 1.923e-06, @@ -13810,7 +13812,8 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", - "output_cost_per_token": 1.923e-06 + "output_cost_per_token": 1.923e-06, + "rpm": 300 }, "cloudflare/@hf/thebloke/codellama-7b-instruct-awq": { "input_cost_per_token": 1.923e-06, @@ -13819,7 +13822,8 @@ "max_output_tokens": 4096, "max_tokens": 4096, "mode": "chat", - "output_cost_per_token": 1.923e-06 + "output_cost_per_token": 1.923e-06, + "rpm": 300 }, "cloudflare/@cf/openai/gpt-oss-120b": { "input_cost_per_token": 3.5e-07, @@ -13829,6 +13833,7 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 7.5e-07, + "rpm": 300, "supports_function_calling": true, "supports_reasoning": true }, @@ -13839,7 +13844,8 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", - "output_cost_per_token": 0.0 + "output_cost_per_token": 0.0, + "rpm": 300 }, "cloudflare/@cf/meta/llama-3.2-3b-instruct": { "input_cost_per_token": 5.09e-08, @@ -13848,7 +13854,8 @@ "max_output_tokens": 80000, "max_tokens": 80000, "mode": "chat", - "output_cost_per_token": 3.35e-07 + "output_cost_per_token": 3.35e-07, + "rpm": 300 }, "cloudflare/@cf/meta/llama-guard-3-8b": { "input_cost_per_token": 4.84e-07, @@ -13857,7 +13864,8 @@ "max_output_tokens": 131072, "max_tokens": 131072, "mode": "chat", - "output_cost_per_token": 3e-08 + "output_cost_per_token": 3e-08, + "rpm": 300 }, "cloudflare/@cf/mistral/mistral-7b-instruct-v0.2-lora": { "input_cost_per_token": 0.0, @@ -13866,7 +13874,8 @@ "max_output_tokens": 15000, "max_tokens": 15000, "mode": "chat", - "output_cost_per_token": 0.0 + "output_cost_per_token": 0.0, + "rpm": 300 }, "cloudflare/@cf/moonshotai/kimi-k2.7-code": { "cache_read_input_token_cost": 1.9e-07, @@ -13877,6 +13886,7 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 4e-06, + "rpm": 20, "supports_function_calling": true, "supports_reasoning": true }, @@ -13888,6 +13898,7 @@ "max_tokens": 80000, "mode": "chat", "output_cost_per_token": 4.881e-06, + "rpm": 300, "supports_reasoning": true }, "cloudflare/@cf/meta/llama-3.1-8b-instruct-fp8": { @@ -13897,7 +13908,8 @@ "max_output_tokens": 32000, "max_tokens": 32000, "mode": "chat", - "output_cost_per_token": 2.87e-07 + "output_cost_per_token": 2.87e-07, + "rpm": 300 }, "cloudflare/@cf/meta/llama-3.2-1b-instruct": { "input_cost_per_token": 2.7e-08, @@ -13906,7 +13918,8 @@ "max_output_tokens": 60000, "max_tokens": 60000, "mode": "chat", - "output_cost_per_token": 2.01e-07 + "output_cost_per_token": 2.01e-07, + "rpm": 300 }, "cloudflare/@cf/moonshotai/kimi-k2.6": { "cache_read_input_token_cost": 1.6e-07, @@ -13917,6 +13930,7 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 4e-06, + "rpm": 20, "supports_function_calling": true, "supports_reasoning": true }, @@ -13928,6 +13942,7 @@ "max_tokens": 131072, "mode": "chat", "output_cost_per_token": 4e-07, + "rpm": 300, "supports_function_calling": true, "supports_reasoning": true }, @@ -13938,7 +13953,8 @@ "max_output_tokens": 8192, "max_tokens": 8192, "mode": "chat", - "output_cost_per_token": 0.0 + "output_cost_per_token": 0.0, + "rpm": 300 }, "cloudflare/@cf/meta/llama-3.3-70b-instruct-fp8-fast": { "input_cost_per_token": 2.93e-07, @@ -13948,6 +13964,7 @@ "max_tokens": 24000, "mode": "chat", "output_cost_per_token": 2.253e-06, + "rpm": 300, "supports_function_calling": true }, "cloudflare/@cf/ibm-granite/granite-4.0-h-micro": { @@ -13958,6 +13975,7 @@ "max_tokens": 131000, "mode": "chat", "output_cost_per_token": 1.12e-07, + "rpm": 300, "supports_function_calling": true }, "cloudflare/@cf/qwen/qwen2.5-coder-32b-instruct": { @@ -13967,7 +13985,8 @@ "max_output_tokens": 32768, "max_tokens": 32768, "mode": "chat", - "output_cost_per_token": 1e-06 + "output_cost_per_token": 1e-06, + "rpm": 300 }, "cloudflare/@cf/zai-org/glm-5.2": { "cache_read_input_token_cost": 2.6e-07, @@ -13978,6 +13997,7 @@ "max_tokens": 262144, "mode": "chat", "output_cost_per_token": 4.4e-06, + "rpm": 20, "supports_function_calling": true, "supports_reasoning": true }, @@ -13989,6 +14009,7 @@ "max_tokens": 256000, "mode": "chat", "output_cost_per_token": 1.5e-06, + "rpm": 300, "supports_function_calling": true, "supports_reasoning": true }, @@ -13999,7 +14020,8 @@ "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", - "output_cost_per_token": 5.55e-07 + "output_cost_per_token": 5.55e-07, + "rpm": 300 }, "cloudflare/@cf/qwen/qwen3-30b-a3b-fp8": { "input_cost_per_token": 5.09e-08, @@ -14009,6 +14031,7 @@ "max_tokens": 32768, "mode": "chat", "output_cost_per_token": 3.35e-07, + "rpm": 300, "supports_function_calling": true, "supports_reasoning": true }, @@ -14019,7 +14042,8 @@ "max_output_tokens": 3500, "max_tokens": 3500, "mode": "chat", - "output_cost_per_token": 0.0 + "output_cost_per_token": 0.0, + "rpm": 300 }, "cloudflare/@cf/google/gemma-4-26b-a4b-it": { "input_cost_per_token": 1e-07, @@ -14029,6 +14053,7 @@ "max_tokens": 256000, "mode": "chat", "output_cost_per_token": 3e-07, + "rpm": 300, "supports_function_calling": true, "supports_reasoning": true }, @@ -14040,6 +14065,7 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 5.55e-07, + "rpm": 300, "supports_function_calling": true }, "cloudflare/@cf/meta/llama-3.2-11b-vision-instruct": { @@ -14050,6 +14076,7 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 6.76e-07, + "rpm": 300, "supports_vision": true }, "cloudflare/@cf/openai/gpt-oss-20b": { @@ -14060,6 +14087,7 @@ "max_tokens": 128000, "mode": "chat", "output_cost_per_token": 3e-07, + "rpm": 300, "supports_function_calling": true, "supports_reasoning": true }, @@ -14071,6 +14099,7 @@ "max_tokens": 131000, "mode": "chat", "output_cost_per_token": 8.5e-07, + "rpm": 300, "supports_function_calling": true }, "cloudflare/@cf/qwen/qwq-32b": { @@ -14081,6 +14110,7 @@ "max_tokens": 24000, "mode": "chat", "output_cost_per_token": 1e-06, + "rpm": 300, "supports_reasoning": true }, "codestral/codestral-2405": { @@ -54901,6 +54931,39 @@ "supports_tool_choice": true, "supports_vision": true }, + "bedrock_mantle/openai.gpt-daybreak-blue-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, + "litellm_provider": "bedrock_mantle", + "max_input_tokens": 1050000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "responses", + "use_openai_responses_path": true, + "supported_endpoints": [ + "/v1/responses" + ], + "supported_modalities": [ + "text", + "image" + ], + "supported_output_modalities": [ + "text" + ], + "supports_function_calling": true, + "supports_prompt_caching": 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-openai-gpt-daybreak-blue-56-sol.html" + }, "bedrock_mantle/openai.gpt-5.6-luna": { "input_cost_per_token": 2.2e-07, "input_cost_per_token_above_272k_tokens": 4.4e-07, @@ -60822,6 +60885,7 @@ "litellm_provider": "cloudflare", "mode": "audio_transcription", "output_cost_per_second": 0.0, + "rpm": 720, "source": "https://developers.cloudflare.com/workers-ai/models/whisper/", "supported_endpoints": [ "/v1/audio/transcriptions" @@ -60832,6 +60896,7 @@ "litellm_provider": "cloudflare", "mode": "audio_transcription", "output_cost_per_second": 0.0, + "rpm": 720, "source": "https://developers.cloudflare.com/workers-ai/models/whisper-large-v3-turbo/", "supported_endpoints": [ "/v1/audio/transcriptions" diff --git a/tests/test_litellm/test_utils.py b/tests/test_litellm/test_utils.py index 8a56a84ade7..cfdbbcb1132 100644 --- a/tests/test_litellm/test_utils.py +++ b/tests/test_litellm/test_utils.py @@ -54,6 +54,12 @@ from litellm.utils import ( # Adds the parent directory to the system path +def test_cloudflare_model_info_includes_rpm(local_model_cost_map): + assert litellm.get_model_info("cloudflare/@cf/meta/llama-3.1-8b-instruct-fp8")["rpm"] == 300 + assert litellm.get_model_info("cloudflare/@cf/moonshotai/kimi-k2.6")["rpm"] == 20 + assert litellm.get_model_info("cloudflare/@cf/openai/whisper-large-v3-turbo")["rpm"] == 720 + + def test_get_utc_datetime_returns_current_aware_utc_time() -> None: before: Final = datetime.now(timezone.utc) result: Final = litellm.utils.get_utc_datetime() From fd1fad5e0559db253842fdb3972a94081c58734d Mon Sep 17 00:00:00 2001 From: mateo Date: Tue, 8 Sep 2026 13:37:56 +0000 Subject: [PATCH 09/11] test: annotate registry metadata test parameters Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- .../responses/test_chatgpt_responses_transformation.py | 6 ++++-- tests/test_litellm/test_utils.py | 2 +- 2 files changed, 5 insertions(+), 3 deletions(-) diff --git a/tests/test_litellm/llms/chatgpt/responses/test_chatgpt_responses_transformation.py b/tests/test_litellm/llms/chatgpt/responses/test_chatgpt_responses_transformation.py index 8a0027fdb53..a7520bd5955 100644 --- a/tests/test_litellm/llms/chatgpt/responses/test_chatgpt_responses_transformation.py +++ b/tests/test_litellm/llms/chatgpt/responses/test_chatgpt_responses_transformation.py @@ -54,7 +54,7 @@ class TestChatGPTResponsesAPITransformation: "chatgpt/gpt-5.6-terra", ], ) - def test_chatgpt_responses_model_metadata(self, model_name, local_model_cost_map): + def test_chatgpt_responses_model_metadata(self, model_name: str, local_model_cost_map: None) -> None: model_info = litellm.get_model_info(model_name) assert model_info["litellm_provider"] == "chatgpt" @@ -75,7 +75,9 @@ class TestChatGPTResponsesAPITransformation: "gpt-5.6-terra", ], ) - def test_chatgpt_models_bridge_chat_completions_to_responses(self, model_name, local_model_cost_map): + def test_chatgpt_models_bridge_chat_completions_to_responses( + self, model_name: str, local_model_cost_map: None + ) -> None: """A chat completions request for these models must take the Responses bridge. `gpt-5.6-*` also exists as an openai chat model, so an unregistered diff --git a/tests/test_litellm/test_utils.py b/tests/test_litellm/test_utils.py index cfdbbcb1132..45532f91f16 100644 --- a/tests/test_litellm/test_utils.py +++ b/tests/test_litellm/test_utils.py @@ -54,7 +54,7 @@ from litellm.utils import ( # Adds the parent directory to the system path -def test_cloudflare_model_info_includes_rpm(local_model_cost_map): +def test_cloudflare_model_info_includes_rpm(local_model_cost_map: None) -> None: assert litellm.get_model_info("cloudflare/@cf/meta/llama-3.1-8b-instruct-fp8")["rpm"] == 300 assert litellm.get_model_info("cloudflare/@cf/moonshotai/kimi-k2.6")["rpm"] == 20 assert litellm.get_model_info("cloudflare/@cf/openai/whisper-large-v3-turbo")["rpm"] == 720 From d497434680186367ce42795ae1fca314503cd2aa Mon Sep 17 00:00:00 2001 From: mateo Date: Tue, 8 Sep 2026 19:28:04 +0000 Subject: [PATCH 10/11] fix(model_prices): update provider and web search pricing Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- ...odel_prices_and_context_window_backup.json | 140 +++++++++++++++--- model_prices_and_context_window.json | 140 +++++++++++++++--- .../test_tool_call_cost_tracking.py | 65 +++++++- ...tex_ai_gemini_transcribe_transformation.py | 4 +- 4 files changed, 293 insertions(+), 56 deletions(-) diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index ef8da798922..a540df1c7d1 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -29176,7 +29176,13 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true + "supports_vision": true, + "search_context_cost_per_query": { + "search_context_size_high": 0.025, + "search_context_size_low": 0.025, + "search_context_size_medium": 0.025 + }, + "supports_web_search": true }, "gpt-4o-mini-2024-07-18": { "cache_read_input_token_cost": 7.5e-08, @@ -29193,9 +29199,9 @@ "output_cost_per_token_priority": 1e-06, "output_cost_per_token_batches": 3e-07, "search_context_cost_per_query": { - "search_context_size_high": 0.03, + "search_context_size_high": 0.025, "search_context_size_low": 0.025, - "search_context_size_medium": 0.0275 + "search_context_size_medium": 0.025 }, "supports_function_calling": true, "supports_parallel_function_calling": true, @@ -29204,7 +29210,8 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true + "supports_vision": true, + "supports_web_search": true }, "gpt-4o-mini-audio-preview": { "deprecation_date": "2026-05-07", @@ -29294,9 +29301,9 @@ "output_cost_per_token": 6e-07, "output_cost_per_token_batches": 3e-07, "search_context_cost_per_query": { - "search_context_size_high": 0.03, + "search_context_size_high": 0.025, "search_context_size_low": 0.025, - "search_context_size_medium": 0.0275 + "search_context_size_medium": 0.025 }, "supports_function_calling": true, "supports_parallel_function_calling": true, @@ -29321,9 +29328,9 @@ "output_cost_per_token": 6e-07, "output_cost_per_token_batches": 3e-07, "search_context_cost_per_query": { - "search_context_size_high": 0.03, + "search_context_size_high": 0.025, "search_context_size_low": 0.025, - "search_context_size_medium": 0.0275 + "search_context_size_medium": 0.025 }, "supports_function_calling": true, "supports_parallel_function_calling": true, @@ -29434,9 +29441,9 @@ "output_cost_per_token": 1e-05, "output_cost_per_token_batches": 5e-06, "search_context_cost_per_query": { - "search_context_size_high": 0.05, - "search_context_size_low": 0.03, - "search_context_size_medium": 0.035 + "search_context_size_high": 0.025, + "search_context_size_low": 0.025, + "search_context_size_medium": 0.025 }, "supports_function_calling": true, "supports_parallel_function_calling": true, @@ -29461,9 +29468,9 @@ "output_cost_per_token": 1e-05, "output_cost_per_token_batches": 5e-06, "search_context_cost_per_query": { - "search_context_size_high": 0.05, - "search_context_size_low": 0.03, - "search_context_size_medium": 0.035 + "search_context_size_high": 0.025, + "search_context_size_low": 0.025, + "search_context_size_medium": 0.025 }, "supports_function_calling": true, "supports_parallel_function_calling": true, @@ -41625,6 +41632,28 @@ "mode": "rerank", "output_cost_per_token": 0.0 }, + "rerank-v4.0-fast": { + "input_cost_per_query": 0.002, + "input_cost_per_token": 0.0, + "litellm_provider": "cohere", + "max_input_tokens": 32768, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "rerank", + "output_cost_per_token": 0.0, + "source": "https://cohere.com/pricing" + }, + "rerank-v4.0-pro": { + "input_cost_per_query": 0.0025, + "input_cost_per_token": 0.0, + "litellm_provider": "cohere", + "max_input_tokens": 32768, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "rerank", + "output_cost_per_token": 0.0, + "source": "https://cohere.com/pricing" + }, "nvidia_nim/nvidia/nv-rerankqa-mistral-4b-v3": { "input_cost_per_query": 0.0, "input_cost_per_token": 0.0, @@ -47811,7 +47840,7 @@ "cache_read_input_token_cost": 5e-08, "input_cost_per_token": 2e-07, "litellm_provider": "vertex_ai", - "max_input_tokens": 2000000, + "max_input_tokens": 128000, "max_output_tokens": 2000000, "max_tokens": 2000000, "mode": "chat", @@ -47827,7 +47856,7 @@ "cache_read_input_token_cost": 5e-08, "input_cost_per_token": 2e-07, "litellm_provider": "vertex_ai", - "max_input_tokens": 2000000, + "max_input_tokens": 128000, "max_output_tokens": 2000000, "max_tokens": 2000000, "mode": "chat", @@ -47842,14 +47871,17 @@ }, "vertex_ai/xai/grok-4.20-non-reasoning": { "cache_read_input_token_cost": 2e-07, - "input_cost_per_token": 2e-06, + "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": "vertex_ai", "max_input_tokens": 2000000, "max_output_tokens": 2000000, "max_tokens": 2000000, "mode": "chat", - "output_cost_per_token": 6e-06, - "source": "https://docs.x.ai/developers/models", + "output_cost_per_token": 2.5e-06, + "output_cost_per_token_above_200k_tokens": 5e-06, + "source": "https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing", "supports_function_calling": true, "supports_response_schema": true, "supports_tool_choice": true, @@ -47858,14 +47890,17 @@ }, "vertex_ai/xai/grok-4.20-reasoning": { "cache_read_input_token_cost": 2e-07, - "input_cost_per_token": 2e-06, + "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": "vertex_ai", "max_input_tokens": 2000000, "max_output_tokens": 2000000, "max_tokens": 2000000, "mode": "chat", - "output_cost_per_token": 6e-06, - "source": "https://docs.x.ai/developers/models", + "output_cost_per_token": 2.5e-06, + "output_cost_per_token_above_200k_tokens": 5e-06, + "source": "https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing", "supports_function_calling": true, "supports_reasoning": true, "supports_response_schema": true, @@ -47873,6 +47908,42 @@ "supports_vision": true, "supports_web_search": true }, + "vertex_ai/xai/grok-4.3": { + "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": "vertex_ai", + "max_input_tokens": 200000, + "max_tokens": 200000, + "mode": "chat", + "output_cost_per_token": 2.5e-06, + "output_cost_per_token_above_200k_tokens": 5e-06, + "source": "https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "vertex_ai/xai/grok-4.6": { + "cache_read_input_token_cost": 5e-07, + "cache_read_input_token_cost_above_200k_tokens": 1e-06, + "input_cost_per_token": 2e-06, + "input_cost_per_token_above_200k_tokens": 4e-06, + "litellm_provider": "vertex_ai", + "max_input_tokens": 524288, + "max_tokens": 524288, + "mode": "chat", + "output_cost_per_token": 6e-06, + "output_cost_per_token_above_200k_tokens": 1.2e-05, + "source": "https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true + }, "vertex_ai/qwen/qwen3-235b-a22b-instruct-2507-maas": { "input_cost_per_token": 2.2e-07, "litellm_provider": "vertex_ai-qwen_models", @@ -56824,8 +56895,8 @@ "rpm": 10 }, "vertex_ai/gemini-3.5-transcribe-preview": { - "input_cost_per_audio_token": 2.5e-06, - "input_cost_per_token": 2.5e-06, + "input_cost_per_audio_token": 2e-06, + "input_cost_per_token": 2e-06, "litellm_provider": "vertex_ai", "mode": "audio_transcription", "output_cost_per_token": 1.2e-05, @@ -56860,6 +56931,27 @@ ], "supports_audio_input": true }, + "vertex_ai/gemini-3.5-live-translate-preview": { + "input_cost_per_audio_token": 3.5e-06, + "input_cost_per_token": 3.5e-06, + "litellm_provider": "vertex_ai", + "mode": "realtime", + "output_cost_per_audio_token": 2.1e-05, + "output_cost_per_token": 2.1e-05, + "source": "https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing", + "supported_endpoints": [ + "/v1/realtime" + ], + "supported_modalities": [ + "audio" + ], + "supported_output_modalities": [ + "audio", + "text" + ], + "supports_audio_input": true, + "supports_audio_output": true + }, "perplexity/pplx-embed-context-v1-0.6b": { "input_cost_per_token": 8e-09, "litellm_provider": "perplexity", diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index ef8da798922..a540df1c7d1 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -29176,7 +29176,13 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true + "supports_vision": true, + "search_context_cost_per_query": { + "search_context_size_high": 0.025, + "search_context_size_low": 0.025, + "search_context_size_medium": 0.025 + }, + "supports_web_search": true }, "gpt-4o-mini-2024-07-18": { "cache_read_input_token_cost": 7.5e-08, @@ -29193,9 +29199,9 @@ "output_cost_per_token_priority": 1e-06, "output_cost_per_token_batches": 3e-07, "search_context_cost_per_query": { - "search_context_size_high": 0.03, + "search_context_size_high": 0.025, "search_context_size_low": 0.025, - "search_context_size_medium": 0.0275 + "search_context_size_medium": 0.025 }, "supports_function_calling": true, "supports_parallel_function_calling": true, @@ -29204,7 +29210,8 @@ "supports_response_schema": true, "supports_system_messages": true, "supports_tool_choice": true, - "supports_vision": true + "supports_vision": true, + "supports_web_search": true }, "gpt-4o-mini-audio-preview": { "deprecation_date": "2026-05-07", @@ -29294,9 +29301,9 @@ "output_cost_per_token": 6e-07, "output_cost_per_token_batches": 3e-07, "search_context_cost_per_query": { - "search_context_size_high": 0.03, + "search_context_size_high": 0.025, "search_context_size_low": 0.025, - "search_context_size_medium": 0.0275 + "search_context_size_medium": 0.025 }, "supports_function_calling": true, "supports_parallel_function_calling": true, @@ -29321,9 +29328,9 @@ "output_cost_per_token": 6e-07, "output_cost_per_token_batches": 3e-07, "search_context_cost_per_query": { - "search_context_size_high": 0.03, + "search_context_size_high": 0.025, "search_context_size_low": 0.025, - "search_context_size_medium": 0.0275 + "search_context_size_medium": 0.025 }, "supports_function_calling": true, "supports_parallel_function_calling": true, @@ -29434,9 +29441,9 @@ "output_cost_per_token": 1e-05, "output_cost_per_token_batches": 5e-06, "search_context_cost_per_query": { - "search_context_size_high": 0.05, - "search_context_size_low": 0.03, - "search_context_size_medium": 0.035 + "search_context_size_high": 0.025, + "search_context_size_low": 0.025, + "search_context_size_medium": 0.025 }, "supports_function_calling": true, "supports_parallel_function_calling": true, @@ -29461,9 +29468,9 @@ "output_cost_per_token": 1e-05, "output_cost_per_token_batches": 5e-06, "search_context_cost_per_query": { - "search_context_size_high": 0.05, - "search_context_size_low": 0.03, - "search_context_size_medium": 0.035 + "search_context_size_high": 0.025, + "search_context_size_low": 0.025, + "search_context_size_medium": 0.025 }, "supports_function_calling": true, "supports_parallel_function_calling": true, @@ -41625,6 +41632,28 @@ "mode": "rerank", "output_cost_per_token": 0.0 }, + "rerank-v4.0-fast": { + "input_cost_per_query": 0.002, + "input_cost_per_token": 0.0, + "litellm_provider": "cohere", + "max_input_tokens": 32768, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "rerank", + "output_cost_per_token": 0.0, + "source": "https://cohere.com/pricing" + }, + "rerank-v4.0-pro": { + "input_cost_per_query": 0.0025, + "input_cost_per_token": 0.0, + "litellm_provider": "cohere", + "max_input_tokens": 32768, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "rerank", + "output_cost_per_token": 0.0, + "source": "https://cohere.com/pricing" + }, "nvidia_nim/nvidia/nv-rerankqa-mistral-4b-v3": { "input_cost_per_query": 0.0, "input_cost_per_token": 0.0, @@ -47811,7 +47840,7 @@ "cache_read_input_token_cost": 5e-08, "input_cost_per_token": 2e-07, "litellm_provider": "vertex_ai", - "max_input_tokens": 2000000, + "max_input_tokens": 128000, "max_output_tokens": 2000000, "max_tokens": 2000000, "mode": "chat", @@ -47827,7 +47856,7 @@ "cache_read_input_token_cost": 5e-08, "input_cost_per_token": 2e-07, "litellm_provider": "vertex_ai", - "max_input_tokens": 2000000, + "max_input_tokens": 128000, "max_output_tokens": 2000000, "max_tokens": 2000000, "mode": "chat", @@ -47842,14 +47871,17 @@ }, "vertex_ai/xai/grok-4.20-non-reasoning": { "cache_read_input_token_cost": 2e-07, - "input_cost_per_token": 2e-06, + "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": "vertex_ai", "max_input_tokens": 2000000, "max_output_tokens": 2000000, "max_tokens": 2000000, "mode": "chat", - "output_cost_per_token": 6e-06, - "source": "https://docs.x.ai/developers/models", + "output_cost_per_token": 2.5e-06, + "output_cost_per_token_above_200k_tokens": 5e-06, + "source": "https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing", "supports_function_calling": true, "supports_response_schema": true, "supports_tool_choice": true, @@ -47858,14 +47890,17 @@ }, "vertex_ai/xai/grok-4.20-reasoning": { "cache_read_input_token_cost": 2e-07, - "input_cost_per_token": 2e-06, + "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": "vertex_ai", "max_input_tokens": 2000000, "max_output_tokens": 2000000, "max_tokens": 2000000, "mode": "chat", - "output_cost_per_token": 6e-06, - "source": "https://docs.x.ai/developers/models", + "output_cost_per_token": 2.5e-06, + "output_cost_per_token_above_200k_tokens": 5e-06, + "source": "https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing", "supports_function_calling": true, "supports_reasoning": true, "supports_response_schema": true, @@ -47873,6 +47908,42 @@ "supports_vision": true, "supports_web_search": true }, + "vertex_ai/xai/grok-4.3": { + "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": "vertex_ai", + "max_input_tokens": 200000, + "max_tokens": 200000, + "mode": "chat", + "output_cost_per_token": 2.5e-06, + "output_cost_per_token_above_200k_tokens": 5e-06, + "source": "https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true + }, + "vertex_ai/xai/grok-4.6": { + "cache_read_input_token_cost": 5e-07, + "cache_read_input_token_cost_above_200k_tokens": 1e-06, + "input_cost_per_token": 2e-06, + "input_cost_per_token_above_200k_tokens": 4e-06, + "litellm_provider": "vertex_ai", + "max_input_tokens": 524288, + "max_tokens": 524288, + "mode": "chat", + "output_cost_per_token": 6e-06, + "output_cost_per_token_above_200k_tokens": 1.2e-05, + "source": "https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true + }, "vertex_ai/qwen/qwen3-235b-a22b-instruct-2507-maas": { "input_cost_per_token": 2.2e-07, "litellm_provider": "vertex_ai-qwen_models", @@ -56824,8 +56895,8 @@ "rpm": 10 }, "vertex_ai/gemini-3.5-transcribe-preview": { - "input_cost_per_audio_token": 2.5e-06, - "input_cost_per_token": 2.5e-06, + "input_cost_per_audio_token": 2e-06, + "input_cost_per_token": 2e-06, "litellm_provider": "vertex_ai", "mode": "audio_transcription", "output_cost_per_token": 1.2e-05, @@ -56860,6 +56931,27 @@ ], "supports_audio_input": true }, + "vertex_ai/gemini-3.5-live-translate-preview": { + "input_cost_per_audio_token": 3.5e-06, + "input_cost_per_token": 3.5e-06, + "litellm_provider": "vertex_ai", + "mode": "realtime", + "output_cost_per_audio_token": 2.1e-05, + "output_cost_per_token": 2.1e-05, + "source": "https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing", + "supported_endpoints": [ + "/v1/realtime" + ], + "supported_modalities": [ + "audio" + ], + "supported_output_modalities": [ + "audio", + "text" + ], + "supports_audio_input": true, + "supports_audio_output": true + }, "perplexity/pplx-embed-context-v1-0.6b": { "input_cost_per_token": 8e-09, "litellm_provider": "perplexity", diff --git a/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_tool_call_cost_tracking.py b/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_tool_call_cost_tracking.py index 6cc3dcceebc..38bf8c544c7 100644 --- a/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_tool_call_cost_tracking.py +++ b/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_tool_call_cost_tracking.py @@ -1,5 +1,6 @@ -import os +import json from collections.abc import Mapping, Sequence +from pathlib import Path import pytest @@ -11,8 +12,6 @@ from litellm.types.llms.openai import FileSearchTool, ResponsesAPIResponse, WebS from litellm.types.utils import ModelResponse, StandardBuiltInToolsParams - - def test_web_search_cost_low(): web_search_options = WebSearchOptions(search_context_size="low") model_info = litellm.get_model_info("gpt-4o-search-preview") @@ -683,12 +682,13 @@ def test_web_search_provider_prefix_fallback_does_not_misprice_non_gemini_model( def _openai_responses_with_web_search_calls(model, num_calls): - from litellm.types.llms.openai import ResponsesAPIResponse from openai.types.responses.response_function_web_search import ( ActionSearch, ResponseFunctionWebSearch, ) + from litellm.types.llms.openai import ResponsesAPIResponse + output = [ ResponseFunctionWebSearch( id=f"ws_{i}", @@ -859,11 +859,64 @@ def test_dated_search_preview_entries_carry_search_pricing(local_model_cost_map) custom_llm_provider="openai", standard_built_in_tools_params=None, ) - assert cost == pytest.approx(0.035), ( - f"dated search-preview id must bill the $0.035 search fee, got ${cost}" + assert cost == pytest.approx(0.025), ( + f"dated search-preview id must bill the $0.025 search fee, got ${cost}" ) +@pytest.mark.parametrize( + "web_search_options", + [ + None, + WebSearchOptions(search_context_size="low"), + WebSearchOptions(search_context_size="medium"), + WebSearchOptions(search_context_size="high"), + ], +) +def test_gpt_4o_mini_snapshot_bills_web_search_like_its_alias( + web_search_options, local_model_cost_map +): + """Snapshot and alias must bill web search identically: OpenAI lists preview search at $25 per 1k calls.""" + alias_info = litellm.get_model_info("gpt-4o-mini") + snapshot_info = litellm.get_model_info("gpt-4o-mini-2024-07-18") + + assert snapshot_info["supports_web_search"] is True + assert alias_info["supports_web_search"] is True + + snapshot_cost = StandardBuiltInToolCostTracking.get_cost_for_web_search( + web_search_options=web_search_options, model_info=snapshot_info + ) + alias_cost = StandardBuiltInToolCostTracking.get_cost_for_web_search( + web_search_options=web_search_options, model_info=alias_info + ) + + assert snapshot_cost == alias_cost == 0.025 + + +def test_gpt_4o_mini_web_search_price_matches_in_both_cost_maps(): + """The bundled backup and canonical file are both served to deployments, so both must carry the price.""" + repo_root = Path(__file__).parents[4] + cost_maps = tuple( + json.loads((repo_root / path).read_text(encoding="utf-8")) + for path in ( + "model_prices_and_context_window.json", + "litellm/model_prices_and_context_window_backup.json", + ) + ) + canonical, backup = cost_maps + expected_search_price = { + "search_context_size_low": 0.025, + "search_context_size_medium": 0.025, + "search_context_size_high": 0.025, + } + for model_name in ("gpt-4o-mini", "gpt-4o-mini-2024-07-18"): + canonical_entry = canonical[model_name] + backup_entry = backup[model_name] + assert canonical_entry["search_context_cost_per_query"] == expected_search_price + assert backup_entry["search_context_cost_per_query"] == expected_search_price + assert canonical_entry == backup_entry + + # Note: File search integration test removed due to complex annotation detection logic # The unit tests in test_azure_assistant_cost_tracking.py provide comprehensive coverage diff --git a/tests/test_litellm/llms/vertex_ai/audio_transcription/test_vertex_ai_gemini_transcribe_transformation.py b/tests/test_litellm/llms/vertex_ai/audio_transcription/test_vertex_ai_gemini_transcribe_transformation.py index eadd87d9c92..08e46b1ffac 100644 --- a/tests/test_litellm/llms/vertex_ai/audio_transcription/test_vertex_ai_gemini_transcribe_transformation.py +++ b/tests/test_litellm/llms/vertex_ai/audio_transcription/test_vertex_ai_gemini_transcribe_transformation.py @@ -322,8 +322,8 @@ class TestModelCostEntry: entry = json.load(f)["vertex_ai/gemini-3.5-transcribe-preview"] assert entry["mode"] == "audio_transcription" assert entry["litellm_provider"] == "vertex_ai" - assert entry["input_cost_per_audio_token"] == pytest.approx(2.5e-06) - assert entry["input_cost_per_token"] == pytest.approx(2.5e-06) + assert entry["input_cost_per_audio_token"] == pytest.approx(2e-06) + assert entry["input_cost_per_token"] == pytest.approx(2e-06) assert entry["output_cost_per_token"] == pytest.approx(1.2e-05) assert entry["supported_endpoints"] == ["/v1/audio/transcriptions"] From 8eb6f6d8350dd0ef7f2e5a61b042213e0b6bd319 Mon Sep 17 00:00:00 2001 From: mateo Date: Tue, 8 Sep 2026 19:43:32 +0000 Subject: [PATCH 11/11] test(cost_calc): type web search test params and drop docstrings Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- .../llm_cost_calc/test_tool_call_cost_tracking.py | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) diff --git a/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_tool_call_cost_tracking.py b/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_tool_call_cost_tracking.py index 38bf8c544c7..f319413045f 100644 --- a/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_tool_call_cost_tracking.py +++ b/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_tool_call_cost_tracking.py @@ -874,9 +874,8 @@ def test_dated_search_preview_entries_carry_search_pricing(local_model_cost_map) ], ) def test_gpt_4o_mini_snapshot_bills_web_search_like_its_alias( - web_search_options, local_model_cost_map -): - """Snapshot and alias must bill web search identically: OpenAI lists preview search at $25 per 1k calls.""" + web_search_options: WebSearchOptions | None, local_model_cost_map: None +) -> None: alias_info = litellm.get_model_info("gpt-4o-mini") snapshot_info = litellm.get_model_info("gpt-4o-mini-2024-07-18") @@ -894,7 +893,6 @@ def test_gpt_4o_mini_snapshot_bills_web_search_like_its_alias( def test_gpt_4o_mini_web_search_price_matches_in_both_cost_maps(): - """The bundled backup and canonical file are both served to deployments, so both must carry the price.""" repo_root = Path(__file__).parents[4] cost_maps = tuple( json.loads((repo_root / path).read_text(encoding="utf-8"))