diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index 7cf956edfdd..a6967c12c94 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -35226,16 +35226,16 @@ "source": "https://nebius.com/prices" }, "nebius/google/gemma-3-27b-it": { - "max_tokens": 128000, - "max_input_tokens": 128000, - "max_output_tokens": 128000, - "input_cost_per_token": 6e-08, - "output_cost_per_token": 2e-07, + "max_tokens": 110000, + "max_input_tokens": 110000, + "max_output_tokens": 110000, + "input_cost_per_token": 1e-07, + "output_cost_per_token": 3e-07, "litellm_provider": "nebius", "mode": "chat", "supports_function_calling": true, "supports_vision": true, - "source": "https://nebius.com/prices" + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/google%2Fgemma-3-27b-it" }, "nebius/meta-llama/Llama-3.3-70B-Instruct": { "max_tokens": 128000, @@ -35347,15 +35347,15 @@ "source": "https://nebius.com/prices" }, "nebius/Qwen/Qwen3-32B": { - "max_tokens": 32768, - "max_input_tokens": 32768, - "max_output_tokens": 32768, + "max_tokens": 40960, + "max_input_tokens": 40960, + "max_output_tokens": 40960, "input_cost_per_token": 1e-07, "output_cost_per_token": 3e-07, "litellm_provider": "nebius", "mode": "chat", "supports_function_calling": true, - "source": "https://nebius.com/prices" + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/Qwen%2FQwen3-32B" }, "nebius/Qwen/Qwen3-30B-A3B": { "max_tokens": 32768, @@ -35436,16 +35436,16 @@ "source": "https://nebius.com/prices" }, "nebius/Qwen/Qwen2.5-VL-72B-Instruct": { - "max_tokens": 131072, - "max_input_tokens": 131072, - "max_output_tokens": 131072, - "input_cost_per_token": 1.3e-07, - "output_cost_per_token": 4e-07, + "max_tokens": 32000, + "max_input_tokens": 32000, + "max_output_tokens": 32000, + "input_cost_per_token": 2.5e-07, + "output_cost_per_token": 7.5e-07, "litellm_provider": "nebius", "mode": "chat", "supports_function_calling": true, "supports_vision": true, - "source": "https://nebius.com/prices" + "source": "https://tokenfactory.nebius.com/models/catalog/image2text/Qwen%2FQwen2.5-VL-72B-Instruct" }, "nebius/Qwen/Qwen2-VL-72B-Instruct": { "max_tokens": 131072, @@ -35470,6 +35470,320 @@ "supports_vision": true, "source": "https://nebius.com/prices" }, + "nebius/deepseek-ai/DeepSeek-V4-Flash": { + "max_tokens": 1048576, + "max_input_tokens": 1048576, + "max_output_tokens": 1048576, + "input_cost_per_token": 1.4e-07, + "output_cost_per_token": 2.8e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/deepseek-ai%2FDeepSeek-V4-Flash" + }, + "nebius/deepseek-ai/DeepSeek-V4-Flash-0731": { + "max_tokens": 1024000, + "max_input_tokens": 1024000, + "max_output_tokens": 1024000, + "input_cost_per_token": 1.4e-07, + "output_cost_per_token": 2.8e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/deepseek-ai%2FDeepSeek-V4-Flash-0731" + }, + "nebius/deepseek-ai/DeepSeek-V4-Pro": { + "max_tokens": 1048576, + "max_input_tokens": 1048576, + "max_output_tokens": 1048576, + "input_cost_per_token": 1.75e-06, + "output_cost_per_token": 3.5e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/deepseek-ai%2FDeepSeek-V4-Pro" + }, + "nebius/MiniMaxAI/MiniMax-M2.5": { + "max_tokens": 196608, + "max_input_tokens": 196608, + "max_output_tokens": 196608, + "input_cost_per_token": 3e-07, + "output_cost_per_token": 1.2e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/MiniMaxAI%2FMiniMax-M2.5" + }, + "nebius/MiniMaxAI/MiniMax-M3": { + "max_tokens": 1048576, + "max_input_tokens": 1048576, + "max_output_tokens": 1048576, + "input_cost_per_token": 3e-07, + "output_cost_per_token": 1.2e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_vision": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/MiniMaxAI%2FMiniMax-M3" + }, + "nebius/moonshotai/Kimi-K2.6": { + "max_tokens": 262144, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "input_cost_per_token": 9.5e-07, + "output_cost_per_token": 4e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_vision": true, + "source": "https://tokenfactory.nebius.com/models/catalog/image2text/moonshotai%2FKimi-K2.6" + }, + "nebius/moonshotai/Kimi-K2.7-Code": { + "max_tokens": 262144, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "input_cost_per_token": 9.5e-07, + "output_cost_per_token": 4e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_vision": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/moonshotai%2FKimi-K2.7-Code" + }, + "nebius/moonshotai/Kimi-K3": { + "max_tokens": 1048576, + "max_input_tokens": 1048576, + "max_output_tokens": 1048576, + "input_cost_per_token": 3e-06, + "output_cost_per_token": 1.5e-05, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_vision": true, + "source": "https://tokenfactory.nebius.com/models/catalog/image2text/moonshotai%2FKimi-K3" + }, + "nebius/NousResearch/Hermes-4-405B": { + "max_tokens": 131072, + "max_input_tokens": 131072, + "max_output_tokens": 131072, + "input_cost_per_token": 1e-06, + "output_cost_per_token": 3e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/NousResearch%2FHermes-4-405B" + }, + "nebius/NousResearch/Hermes-4-70B": { + "max_tokens": 131072, + "max_input_tokens": 131072, + "max_output_tokens": 131072, + "input_cost_per_token": 1.3e-07, + "output_cost_per_token": 4e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/NousResearch%2FHermes-4-70B" + }, + "nebius/nvidia/Cosmos3-Super-Reasoner": { + "max_tokens": 262144, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "input_cost_per_token": 1e-07, + "output_cost_per_token": 3e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_vision": true, + "source": "https://tokenfactory.nebius.com/models/catalog/image2text/nvidia%2FCosmos3-Super-Reasoner" + }, + "nebius/nvidia/Llama-3_1-Nemotron-Ultra-253B-v1": { + "max_tokens": 131072, + "max_input_tokens": 131072, + "max_output_tokens": 131072, + "input_cost_per_token": 6e-07, + "output_cost_per_token": 1.8e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/nvidia%2FLlama-3_1-Nemotron-Ultra-253B-v1" + }, + "nebius/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B": { + "max_tokens": 262144, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "input_cost_per_token": 6e-08, + "output_cost_per_token": 2.4e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/nvidia%2FNVIDIA-Nemotron-3-Nano-30B-A3B" + }, + "nebius/nvidia/Nemotron-3-Nano-Omni": { + "max_tokens": 262144, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "input_cost_per_token": 6e-08, + "output_cost_per_token": 2.4e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_vision": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/nvidia%2FNemotron-3-Nano-Omni" + }, + "nebius/nvidia/nemotron-3-super-120b-a12b": { + "max_tokens": 262144, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "input_cost_per_token": 3e-07, + "output_cost_per_token": 9e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/nvidia%2Fnemotron-3-super-120b-a12b" + }, + "nebius/nvidia/Nemotron-3-Ultra-550b-a55b": { + "max_tokens": 1048576, + "max_input_tokens": 1048576, + "max_output_tokens": 1048576, + "input_cost_per_token": 1e-06, + "output_cost_per_token": 3e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/nvidia%2FNemotron-3-Ultra-550b-a55b" + }, + "nebius/nvidia/Nemotron-3_5-Lightning": { + "max_tokens": 1048576, + "max_input_tokens": 1048576, + "max_output_tokens": 1048576, + "input_cost_per_token": 6e-08, + "output_cost_per_token": 2.4e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/nvidia%2FNemotron-3_5-Lightning" + }, + "nebius/openai/gpt-oss-120b": { + "max_tokens": 131072, + "max_input_tokens": 131072, + "max_output_tokens": 131072, + "input_cost_per_token": 1.5e-07, + "output_cost_per_token": 6e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/openai%2Fgpt-oss-120b" + }, + "nebius/openbmb/MiniCPM-V-4_5": { + "max_tokens": 32000, + "max_input_tokens": 32000, + "max_output_tokens": 32000, + "input_cost_per_token": 6.58e-07, + "output_cost_per_token": 1.11e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_vision": true, + "source": "https://tokenfactory.nebius.com/models/catalog/image2text/openbmb%2FMiniCPM-V-4_5" + }, + "nebius/Qwen/Qwen3-235B-A22B-Instruct-2507": { + "max_tokens": 262144, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "input_cost_per_token": 2e-07, + "output_cost_per_token": 6e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/Qwen%2FQwen3-235B-A22B-Instruct-2507" + }, + "nebius/Qwen/Qwen3-30B-A3B-Instruct-2507": { + "max_tokens": 262144, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "input_cost_per_token": 1e-07, + "output_cost_per_token": 3e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/Qwen%2FQwen3-30B-A3B-Instruct-2507" + }, + "nebius/Qwen/Qwen3-Next-80B-A3B-Thinking": { + "max_tokens": 128000, + "max_input_tokens": 128000, + "max_output_tokens": 128000, + "input_cost_per_token": 1.5e-07, + "output_cost_per_token": 1.2e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/Qwen%2FQwen3-Next-80B-A3B-Thinking" + }, + "nebius/Qwen/Qwen3.5-397B-A17B": { + "max_tokens": 262144, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "input_cost_per_token": 6e-07, + "output_cost_per_token": 3.6e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/Qwen%2FQwen3.5-397B-A17B" + }, + "nebius/zai-org/GLM-5.1": { + "max_tokens": 202752, + "max_input_tokens": 202752, + "max_output_tokens": 202752, + "input_cost_per_token": 1.4e-06, + "output_cost_per_token": 4.4e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/zai-org%2FGLM-5.1" + }, + "nebius/zai-org/GLM-5.2": { + "max_tokens": 1048576, + "max_input_tokens": 1048576, + "max_output_tokens": 1048576, + "input_cost_per_token": 1.4e-06, + "output_cost_per_token": 4.4e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/zai-org%2FGLM-5.2" + }, + "nebius/zai-org/GLM-5.3-Flash": { + "max_tokens": 1024000, + "max_input_tokens": 1024000, + "max_output_tokens": 1024000, + "input_cost_per_token": 1.5e-07, + "output_cost_per_token": 5e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/zai-org%2FGLM-5.3-Flash" + }, "nebius/BAAI/bge-en-icl": { "max_tokens": 32768, "max_input_tokens": 32768, @@ -35497,6 +35811,15 @@ "mode": "embedding", "source": "https://nebius.com/prices" }, + "nebius/Qwen/Qwen3-Embedding-8B": { + "max_tokens": 40960, + "max_input_tokens": 40960, + "input_cost_per_token": 1e-08, + "output_cost_per_token": 0.0, + "litellm_provider": "nebius", + "mode": "embedding", + "source": "https://tokenfactory.nebius.com/models/catalog/embedding/Qwen%2FQwen3-Embedding-8B" + }, "nvidia.nemotron-nano-12b-v2": { "input_cost_per_token": 2e-07, "litellm_provider": "bedrock_converse", @@ -46551,16 +46874,30 @@ "supports_vision": false }, "watsonx/bigscience/mt0-xxl-13b": { - "max_tokens": 8192, - "max_input_tokens": 8192, - "max_output_tokens": 8192, - "input_cost_per_token": 0.0005, - "output_cost_per_token": 0.002, + "max_tokens": 4096, + "max_input_tokens": 4096, + "max_output_tokens": 4096, + "input_cost_per_token": 1.908e-06, + "output_cost_per_token": 1.908e-06, "litellm_provider": "watsonx", "mode": "chat", "supports_function_calling": false, "supports_parallel_function_calling": false, - "supports_vision": false + "supports_vision": false, + "source": "https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-models.html?context=wx" + }, + "watsonx/bigscience/mt0-xxl": { + "max_tokens": 4096, + "max_input_tokens": 4096, + "max_output_tokens": 4096, + "input_cost_per_token": 1.908e-06, + "output_cost_per_token": 1.908e-06, + "litellm_provider": "watsonx", + "mode": "chat", + "supports_function_calling": false, + "supports_parallel_function_calling": false, + "supports_vision": false, + "source": "https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-models.html?context=wx" }, "watsonx/core42/jais-13b-chat": { "max_tokens": 8192, @@ -46623,16 +46960,17 @@ "supports_vision": false }, "watsonx/ibm/granite-4-h-small": { - "max_tokens": 20480, - "max_input_tokens": 20480, + "max_tokens": 131072, + "max_input_tokens": 131072, "max_output_tokens": 20480, - "input_cost_per_token": 6e-08, - "output_cost_per_token": 2.5e-07, + "input_cost_per_token": 6.36e-08, + "output_cost_per_token": 2.65e-07, "litellm_provider": "watsonx", "mode": "chat", "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": false + "supports_vision": false, + "source": "https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-models.html?context=wx" }, "watsonx/ibm/granite-guardian-3-2-2b": { "max_tokens": 8192, @@ -46755,28 +47093,43 @@ "supports_vision": true }, "watsonx/meta-llama/llama-3-3-70b-instruct": { - "max_tokens": 128000, - "max_input_tokens": 128000, + "max_tokens": 131072, + "max_input_tokens": 131072, "max_output_tokens": 128000, - "input_cost_per_token": 7.1e-07, - "output_cost_per_token": 7.1e-07, + "input_cost_per_token": 7.526e-07, + "output_cost_per_token": 7.526e-07, "litellm_provider": "watsonx", "mode": "chat", "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": false + "supports_vision": false, + "source": "https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-models.html?context=wx" }, "watsonx/meta-llama/llama-4-maverick-17b": { - "max_tokens": 128000, - "max_input_tokens": 128000, + "max_tokens": 131072, + "max_input_tokens": 131072, "max_output_tokens": 128000, - "input_cost_per_token": 3.5e-07, - "output_cost_per_token": 1.4e-06, + "input_cost_per_token": 3.71e-07, + "output_cost_per_token": 1.484e-06, "litellm_provider": "watsonx", "mode": "chat", "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": false + "supports_vision": false, + "source": "https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-models.html?context=wx" + }, + "watsonx/meta-llama/llama-4-maverick-17b-128e-instruct-fp8": { + "max_tokens": 131072, + "max_input_tokens": 131072, + "max_output_tokens": 128000, + "input_cost_per_token": 3.71e-07, + "output_cost_per_token": 1.484e-06, + "litellm_provider": "watsonx", + "mode": "chat", + "supports_function_calling": true, + "supports_parallel_function_calling": true, + "supports_vision": false, + "source": "https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-models.html?context=wx" }, "watsonx/meta-llama/llama-guard-3-11b-vision": { "max_tokens": 128000, @@ -46815,16 +47168,17 @@ "supports_vision": false }, "watsonx/mistralai/mistral-small-3-1-24b-instruct-2503": { - "max_tokens": 32000, - "max_input_tokens": 32000, + "max_tokens": 131072, + "max_input_tokens": 131072, "max_output_tokens": 32000, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 3e-07, + "input_cost_per_token": 1.06e-07, + "output_cost_per_token": 3.18e-07, "litellm_provider": "watsonx", "mode": "chat", "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": false + "supports_vision": false, + "source": "https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-models.html?context=wx" }, "watsonx/mistralai/pixtral-12b-2409": { "max_tokens": 128000, @@ -46839,16 +47193,17 @@ "supports_vision": true }, "watsonx/openai/gpt-oss-120b": { - "max_tokens": 8192, - "max_input_tokens": 8192, + "max_tokens": 131072, + "max_input_tokens": 131072, "max_output_tokens": 8192, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, + "input_cost_per_token": 1.59e-07, + "output_cost_per_token": 6.36e-07, "litellm_provider": "watsonx", "mode": "chat", "supports_function_calling": false, "supports_parallel_function_calling": false, - "supports_vision": false + "supports_vision": false, + "source": "https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-models.html?context=wx" }, "watsonx/sdaia/allam-1-13b-instruct": { "max_tokens": 8192, @@ -53520,6 +53875,32 @@ } ] }, + "volcengine/doubao-seed-2-1-pro-260628": { + "litellm_provider": "volcengine", + "max_input_tokens": 256000, + "max_output_tokens": 256000, + "max_tokens": 256000, + "mode": "chat", + "source": "https://www.volcengine.com/docs/82379/1330310", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": false, + "supports_vision": true + }, + "volcengine/doubao-seed-2-1-turbo-260628": { + "litellm_provider": "volcengine", + "max_input_tokens": 256000, + "max_output_tokens": 256000, + "max_tokens": 256000, + "mode": "chat", + "source": "https://www.volcengine.com/docs/82379/1330310", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": false, + "supports_vision": true + }, "bedrock/us-east-1/zai.glm-5": { "input_cost_per_token": 1e-06, "output_cost_per_token": 3.2e-06, diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index 7cf956edfdd..a6967c12c94 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -35226,16 +35226,16 @@ "source": "https://nebius.com/prices" }, "nebius/google/gemma-3-27b-it": { - "max_tokens": 128000, - "max_input_tokens": 128000, - "max_output_tokens": 128000, - "input_cost_per_token": 6e-08, - "output_cost_per_token": 2e-07, + "max_tokens": 110000, + "max_input_tokens": 110000, + "max_output_tokens": 110000, + "input_cost_per_token": 1e-07, + "output_cost_per_token": 3e-07, "litellm_provider": "nebius", "mode": "chat", "supports_function_calling": true, "supports_vision": true, - "source": "https://nebius.com/prices" + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/google%2Fgemma-3-27b-it" }, "nebius/meta-llama/Llama-3.3-70B-Instruct": { "max_tokens": 128000, @@ -35347,15 +35347,15 @@ "source": "https://nebius.com/prices" }, "nebius/Qwen/Qwen3-32B": { - "max_tokens": 32768, - "max_input_tokens": 32768, - "max_output_tokens": 32768, + "max_tokens": 40960, + "max_input_tokens": 40960, + "max_output_tokens": 40960, "input_cost_per_token": 1e-07, "output_cost_per_token": 3e-07, "litellm_provider": "nebius", "mode": "chat", "supports_function_calling": true, - "source": "https://nebius.com/prices" + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/Qwen%2FQwen3-32B" }, "nebius/Qwen/Qwen3-30B-A3B": { "max_tokens": 32768, @@ -35436,16 +35436,16 @@ "source": "https://nebius.com/prices" }, "nebius/Qwen/Qwen2.5-VL-72B-Instruct": { - "max_tokens": 131072, - "max_input_tokens": 131072, - "max_output_tokens": 131072, - "input_cost_per_token": 1.3e-07, - "output_cost_per_token": 4e-07, + "max_tokens": 32000, + "max_input_tokens": 32000, + "max_output_tokens": 32000, + "input_cost_per_token": 2.5e-07, + "output_cost_per_token": 7.5e-07, "litellm_provider": "nebius", "mode": "chat", "supports_function_calling": true, "supports_vision": true, - "source": "https://nebius.com/prices" + "source": "https://tokenfactory.nebius.com/models/catalog/image2text/Qwen%2FQwen2.5-VL-72B-Instruct" }, "nebius/Qwen/Qwen2-VL-72B-Instruct": { "max_tokens": 131072, @@ -35470,6 +35470,320 @@ "supports_vision": true, "source": "https://nebius.com/prices" }, + "nebius/deepseek-ai/DeepSeek-V4-Flash": { + "max_tokens": 1048576, + "max_input_tokens": 1048576, + "max_output_tokens": 1048576, + "input_cost_per_token": 1.4e-07, + "output_cost_per_token": 2.8e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/deepseek-ai%2FDeepSeek-V4-Flash" + }, + "nebius/deepseek-ai/DeepSeek-V4-Flash-0731": { + "max_tokens": 1024000, + "max_input_tokens": 1024000, + "max_output_tokens": 1024000, + "input_cost_per_token": 1.4e-07, + "output_cost_per_token": 2.8e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/deepseek-ai%2FDeepSeek-V4-Flash-0731" + }, + "nebius/deepseek-ai/DeepSeek-V4-Pro": { + "max_tokens": 1048576, + "max_input_tokens": 1048576, + "max_output_tokens": 1048576, + "input_cost_per_token": 1.75e-06, + "output_cost_per_token": 3.5e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/deepseek-ai%2FDeepSeek-V4-Pro" + }, + "nebius/MiniMaxAI/MiniMax-M2.5": { + "max_tokens": 196608, + "max_input_tokens": 196608, + "max_output_tokens": 196608, + "input_cost_per_token": 3e-07, + "output_cost_per_token": 1.2e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/MiniMaxAI%2FMiniMax-M2.5" + }, + "nebius/MiniMaxAI/MiniMax-M3": { + "max_tokens": 1048576, + "max_input_tokens": 1048576, + "max_output_tokens": 1048576, + "input_cost_per_token": 3e-07, + "output_cost_per_token": 1.2e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_vision": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/MiniMaxAI%2FMiniMax-M3" + }, + "nebius/moonshotai/Kimi-K2.6": { + "max_tokens": 262144, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "input_cost_per_token": 9.5e-07, + "output_cost_per_token": 4e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_vision": true, + "source": "https://tokenfactory.nebius.com/models/catalog/image2text/moonshotai%2FKimi-K2.6" + }, + "nebius/moonshotai/Kimi-K2.7-Code": { + "max_tokens": 262144, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "input_cost_per_token": 9.5e-07, + "output_cost_per_token": 4e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_vision": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/moonshotai%2FKimi-K2.7-Code" + }, + "nebius/moonshotai/Kimi-K3": { + "max_tokens": 1048576, + "max_input_tokens": 1048576, + "max_output_tokens": 1048576, + "input_cost_per_token": 3e-06, + "output_cost_per_token": 1.5e-05, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_vision": true, + "source": "https://tokenfactory.nebius.com/models/catalog/image2text/moonshotai%2FKimi-K3" + }, + "nebius/NousResearch/Hermes-4-405B": { + "max_tokens": 131072, + "max_input_tokens": 131072, + "max_output_tokens": 131072, + "input_cost_per_token": 1e-06, + "output_cost_per_token": 3e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/NousResearch%2FHermes-4-405B" + }, + "nebius/NousResearch/Hermes-4-70B": { + "max_tokens": 131072, + "max_input_tokens": 131072, + "max_output_tokens": 131072, + "input_cost_per_token": 1.3e-07, + "output_cost_per_token": 4e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/NousResearch%2FHermes-4-70B" + }, + "nebius/nvidia/Cosmos3-Super-Reasoner": { + "max_tokens": 262144, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "input_cost_per_token": 1e-07, + "output_cost_per_token": 3e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_vision": true, + "source": "https://tokenfactory.nebius.com/models/catalog/image2text/nvidia%2FCosmos3-Super-Reasoner" + }, + "nebius/nvidia/Llama-3_1-Nemotron-Ultra-253B-v1": { + "max_tokens": 131072, + "max_input_tokens": 131072, + "max_output_tokens": 131072, + "input_cost_per_token": 6e-07, + "output_cost_per_token": 1.8e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/nvidia%2FLlama-3_1-Nemotron-Ultra-253B-v1" + }, + "nebius/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B": { + "max_tokens": 262144, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "input_cost_per_token": 6e-08, + "output_cost_per_token": 2.4e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/nvidia%2FNVIDIA-Nemotron-3-Nano-30B-A3B" + }, + "nebius/nvidia/Nemotron-3-Nano-Omni": { + "max_tokens": 262144, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "input_cost_per_token": 6e-08, + "output_cost_per_token": 2.4e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_vision": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/nvidia%2FNemotron-3-Nano-Omni" + }, + "nebius/nvidia/nemotron-3-super-120b-a12b": { + "max_tokens": 262144, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "input_cost_per_token": 3e-07, + "output_cost_per_token": 9e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/nvidia%2Fnemotron-3-super-120b-a12b" + }, + "nebius/nvidia/Nemotron-3-Ultra-550b-a55b": { + "max_tokens": 1048576, + "max_input_tokens": 1048576, + "max_output_tokens": 1048576, + "input_cost_per_token": 1e-06, + "output_cost_per_token": 3e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/nvidia%2FNemotron-3-Ultra-550b-a55b" + }, + "nebius/nvidia/Nemotron-3_5-Lightning": { + "max_tokens": 1048576, + "max_input_tokens": 1048576, + "max_output_tokens": 1048576, + "input_cost_per_token": 6e-08, + "output_cost_per_token": 2.4e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/nvidia%2FNemotron-3_5-Lightning" + }, + "nebius/openai/gpt-oss-120b": { + "max_tokens": 131072, + "max_input_tokens": 131072, + "max_output_tokens": 131072, + "input_cost_per_token": 1.5e-07, + "output_cost_per_token": 6e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/openai%2Fgpt-oss-120b" + }, + "nebius/openbmb/MiniCPM-V-4_5": { + "max_tokens": 32000, + "max_input_tokens": 32000, + "max_output_tokens": 32000, + "input_cost_per_token": 6.58e-07, + "output_cost_per_token": 1.11e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_vision": true, + "source": "https://tokenfactory.nebius.com/models/catalog/image2text/openbmb%2FMiniCPM-V-4_5" + }, + "nebius/Qwen/Qwen3-235B-A22B-Instruct-2507": { + "max_tokens": 262144, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "input_cost_per_token": 2e-07, + "output_cost_per_token": 6e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/Qwen%2FQwen3-235B-A22B-Instruct-2507" + }, + "nebius/Qwen/Qwen3-30B-A3B-Instruct-2507": { + "max_tokens": 262144, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "input_cost_per_token": 1e-07, + "output_cost_per_token": 3e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/Qwen%2FQwen3-30B-A3B-Instruct-2507" + }, + "nebius/Qwen/Qwen3-Next-80B-A3B-Thinking": { + "max_tokens": 128000, + "max_input_tokens": 128000, + "max_output_tokens": 128000, + "input_cost_per_token": 1.5e-07, + "output_cost_per_token": 1.2e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/Qwen%2FQwen3-Next-80B-A3B-Thinking" + }, + "nebius/Qwen/Qwen3.5-397B-A17B": { + "max_tokens": 262144, + "max_input_tokens": 262144, + "max_output_tokens": 262144, + "input_cost_per_token": 6e-07, + "output_cost_per_token": 3.6e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/Qwen%2FQwen3.5-397B-A17B" + }, + "nebius/zai-org/GLM-5.1": { + "max_tokens": 202752, + "max_input_tokens": 202752, + "max_output_tokens": 202752, + "input_cost_per_token": 1.4e-06, + "output_cost_per_token": 4.4e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/zai-org%2FGLM-5.1" + }, + "nebius/zai-org/GLM-5.2": { + "max_tokens": 1048576, + "max_input_tokens": 1048576, + "max_output_tokens": 1048576, + "input_cost_per_token": 1.4e-06, + "output_cost_per_token": 4.4e-06, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/zai-org%2FGLM-5.2" + }, + "nebius/zai-org/GLM-5.3-Flash": { + "max_tokens": 1024000, + "max_input_tokens": 1024000, + "max_output_tokens": 1024000, + "input_cost_per_token": 1.5e-07, + "output_cost_per_token": 5e-07, + "litellm_provider": "nebius", + "mode": "chat", + "supports_function_calling": true, + "supports_reasoning": true, + "source": "https://tokenfactory.nebius.com/models/catalog/text2text/zai-org%2FGLM-5.3-Flash" + }, "nebius/BAAI/bge-en-icl": { "max_tokens": 32768, "max_input_tokens": 32768, @@ -35497,6 +35811,15 @@ "mode": "embedding", "source": "https://nebius.com/prices" }, + "nebius/Qwen/Qwen3-Embedding-8B": { + "max_tokens": 40960, + "max_input_tokens": 40960, + "input_cost_per_token": 1e-08, + "output_cost_per_token": 0.0, + "litellm_provider": "nebius", + "mode": "embedding", + "source": "https://tokenfactory.nebius.com/models/catalog/embedding/Qwen%2FQwen3-Embedding-8B" + }, "nvidia.nemotron-nano-12b-v2": { "input_cost_per_token": 2e-07, "litellm_provider": "bedrock_converse", @@ -46551,16 +46874,30 @@ "supports_vision": false }, "watsonx/bigscience/mt0-xxl-13b": { - "max_tokens": 8192, - "max_input_tokens": 8192, - "max_output_tokens": 8192, - "input_cost_per_token": 0.0005, - "output_cost_per_token": 0.002, + "max_tokens": 4096, + "max_input_tokens": 4096, + "max_output_tokens": 4096, + "input_cost_per_token": 1.908e-06, + "output_cost_per_token": 1.908e-06, "litellm_provider": "watsonx", "mode": "chat", "supports_function_calling": false, "supports_parallel_function_calling": false, - "supports_vision": false + "supports_vision": false, + "source": "https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-models.html?context=wx" + }, + "watsonx/bigscience/mt0-xxl": { + "max_tokens": 4096, + "max_input_tokens": 4096, + "max_output_tokens": 4096, + "input_cost_per_token": 1.908e-06, + "output_cost_per_token": 1.908e-06, + "litellm_provider": "watsonx", + "mode": "chat", + "supports_function_calling": false, + "supports_parallel_function_calling": false, + "supports_vision": false, + "source": "https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-models.html?context=wx" }, "watsonx/core42/jais-13b-chat": { "max_tokens": 8192, @@ -46623,16 +46960,17 @@ "supports_vision": false }, "watsonx/ibm/granite-4-h-small": { - "max_tokens": 20480, - "max_input_tokens": 20480, + "max_tokens": 131072, + "max_input_tokens": 131072, "max_output_tokens": 20480, - "input_cost_per_token": 6e-08, - "output_cost_per_token": 2.5e-07, + "input_cost_per_token": 6.36e-08, + "output_cost_per_token": 2.65e-07, "litellm_provider": "watsonx", "mode": "chat", "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": false + "supports_vision": false, + "source": "https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-models.html?context=wx" }, "watsonx/ibm/granite-guardian-3-2-2b": { "max_tokens": 8192, @@ -46755,28 +47093,43 @@ "supports_vision": true }, "watsonx/meta-llama/llama-3-3-70b-instruct": { - "max_tokens": 128000, - "max_input_tokens": 128000, + "max_tokens": 131072, + "max_input_tokens": 131072, "max_output_tokens": 128000, - "input_cost_per_token": 7.1e-07, - "output_cost_per_token": 7.1e-07, + "input_cost_per_token": 7.526e-07, + "output_cost_per_token": 7.526e-07, "litellm_provider": "watsonx", "mode": "chat", "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": false + "supports_vision": false, + "source": "https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-models.html?context=wx" }, "watsonx/meta-llama/llama-4-maverick-17b": { - "max_tokens": 128000, - "max_input_tokens": 128000, + "max_tokens": 131072, + "max_input_tokens": 131072, "max_output_tokens": 128000, - "input_cost_per_token": 3.5e-07, - "output_cost_per_token": 1.4e-06, + "input_cost_per_token": 3.71e-07, + "output_cost_per_token": 1.484e-06, "litellm_provider": "watsonx", "mode": "chat", "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": false + "supports_vision": false, + "source": "https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-models.html?context=wx" + }, + "watsonx/meta-llama/llama-4-maverick-17b-128e-instruct-fp8": { + "max_tokens": 131072, + "max_input_tokens": 131072, + "max_output_tokens": 128000, + "input_cost_per_token": 3.71e-07, + "output_cost_per_token": 1.484e-06, + "litellm_provider": "watsonx", + "mode": "chat", + "supports_function_calling": true, + "supports_parallel_function_calling": true, + "supports_vision": false, + "source": "https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-models.html?context=wx" }, "watsonx/meta-llama/llama-guard-3-11b-vision": { "max_tokens": 128000, @@ -46815,16 +47168,17 @@ "supports_vision": false }, "watsonx/mistralai/mistral-small-3-1-24b-instruct-2503": { - "max_tokens": 32000, - "max_input_tokens": 32000, + "max_tokens": 131072, + "max_input_tokens": 131072, "max_output_tokens": 32000, - "input_cost_per_token": 1e-07, - "output_cost_per_token": 3e-07, + "input_cost_per_token": 1.06e-07, + "output_cost_per_token": 3.18e-07, "litellm_provider": "watsonx", "mode": "chat", "supports_function_calling": true, "supports_parallel_function_calling": true, - "supports_vision": false + "supports_vision": false, + "source": "https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-models.html?context=wx" }, "watsonx/mistralai/pixtral-12b-2409": { "max_tokens": 128000, @@ -46839,16 +47193,17 @@ "supports_vision": true }, "watsonx/openai/gpt-oss-120b": { - "max_tokens": 8192, - "max_input_tokens": 8192, + "max_tokens": 131072, + "max_input_tokens": 131072, "max_output_tokens": 8192, - "input_cost_per_token": 1.5e-07, - "output_cost_per_token": 6e-07, + "input_cost_per_token": 1.59e-07, + "output_cost_per_token": 6.36e-07, "litellm_provider": "watsonx", "mode": "chat", "supports_function_calling": false, "supports_parallel_function_calling": false, - "supports_vision": false + "supports_vision": false, + "source": "https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-models.html?context=wx" }, "watsonx/sdaia/allam-1-13b-instruct": { "max_tokens": 8192, @@ -53520,6 +53875,32 @@ } ] }, + "volcengine/doubao-seed-2-1-pro-260628": { + "litellm_provider": "volcengine", + "max_input_tokens": 256000, + "max_output_tokens": 256000, + "max_tokens": 256000, + "mode": "chat", + "source": "https://www.volcengine.com/docs/82379/1330310", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": false, + "supports_vision": true + }, + "volcengine/doubao-seed-2-1-turbo-260628": { + "litellm_provider": "volcengine", + "max_input_tokens": 256000, + "max_output_tokens": 256000, + "max_tokens": 256000, + "mode": "chat", + "source": "https://www.volcengine.com/docs/82379/1330310", + "supports_function_calling": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": false, + "supports_vision": true + }, "bedrock/us-east-1/zai.glm-5": { "input_cost_per_token": 1e-06, "output_cost_per_token": 3.2e-06,