scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-10-05 02:41:56 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 49

fix(registry): sync Azure o-series and Together deprecation dates, gpt-5.4-mini/nano context window

Browse source 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
This commit is contained in:
mateo 2026-09-12 13:24:00 +00:00
parent 9071ca503e
commit 6c974df9a0
2 changed files with 52 additions and 30 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

41
litellm/model_prices_and_context_window_backup.json
View file

@ -8730,7 +8730,7 @@
"supports_function_calling": true
},
"azure/o1": {
"deprecation_date": "2026-10-21",
"deprecation_date": "2026-11-19",
"cache_read_input_token_cost": 7.5e-06,
"input_cost_per_token": 1.5e-05,
"litellm_provider": "azure",
@ -8748,7 +8748,7 @@
},
"azure/o1-2024-12-17": {
"cache_read_input_token_cost": 7.5e-06,
"deprecation_date": "2026-10-21",
"deprecation_date": "2026-11-19",
"input_cost_per_token": 1.5e-05,
"litellm_provider": "azure",
"max_input_tokens": 200000,
@ -8825,7 +8825,7 @@
"supports_vision": false
},
"azure/o3": {
"deprecation_date": "2026-10-21",
"deprecation_date": "2026-11-19",
"cache_read_input_token_cost": 5e-07,
"input_cost_per_token": 2e-06,
"litellm_provider": "azure",
@ -8855,7 +8855,7 @@
"supports_vision": true
},
"azure/o3-2025-04-16": {
"deprecation_date": "2026-10-21",
"deprecation_date": "2026-11-19",
"cache_read_input_token_cost": 5e-07,
"input_cost_per_token": 2e-06,
"litellm_provider": "azure",
@ -8886,7 +8886,7 @@
},
"azure/o3-deep-research": {
"cache_read_input_token_cost": 2.5e-06,
"deprecation_date": "2026-12-26",
"deprecation_date": "2026-11-19",
"input_cost_per_token": 1e-05,
"litellm_provider": "azure",
"max_input_tokens": 200000,
@ -8923,7 +8923,7 @@
"supports_web_search": true
},
"azure/o3-mini": {
"deprecation_date": "2026-10-01",
"deprecation_date": "2026-11-19",
"cache_read_input_token_cost": 5.5e-07,
"input_cost_per_token": 1.1e-06,
"litellm_provider": "azure",
@ -8940,7 +8940,7 @@
},
"azure/o3-mini-2025-01-31": {
"cache_read_input_token_cost": 5.5e-07,
"deprecation_date": "2026-10-01",
"deprecation_date": "2026-11-19",
"input_cost_per_token": 1.1e-06,
"litellm_provider": "azure",
"max_input_tokens": 200000,
@ -8954,7 +8954,7 @@
"supports_vision": false
},
"azure/o3-pro": {
"deprecation_date": "2026-12-17",
"deprecation_date": "2026-11-19",
"input_cost_per_token": 2e-05,
"input_cost_per_token_batches": 1e-05,
"litellm_provider": "azure",
@ -8985,7 +8985,7 @@
"supports_vision": true
},
"azure/o3-pro-2025-06-10": {
"deprecation_date": "2026-12-17",
"deprecation_date": "2026-11-19",
"input_cost_per_token": 2e-05,
"input_cost_per_token_batches": 1e-05,
"litellm_provider": "azure",
@ -9016,7 +9016,7 @@
"supports_vision": true
},
"azure/o4-mini": {
"deprecation_date": "2026-10-16",
"deprecation_date": "2026-11-19",
"cache_read_input_token_cost": 2.75e-07,
"input_cost_per_token": 1.1e-06,
"litellm_provider": "azure",
@ -9047,7 +9047,7 @@
},
"azure/o4-mini-2025-04-16": {
"cache_read_input_token_cost": 2.75e-07,
"deprecation_date": "2026-10-16",
"deprecation_date": "2026-11-19",
"input_cost_per_token": 1.1e-06,
"litellm_provider": "azure",
"max_input_tokens": 200000,
@ -31650,7 +31650,7 @@
"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": 400000,
"max_output_tokens": 128000,
"max_tokens": 128000,
"mode": "chat",
@ -31702,7 +31702,7 @@
"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": 400000,
"max_output_tokens": 128000,
"max_tokens": 128000,
"mode": "chat",
@ -31752,7 +31752,7 @@
"input_cost_per_token_flex": 1e-07,
"input_cost_per_token_batches": 1e-07,
"litellm_provider": "openai",
"max_input_tokens": 272000,
"max_input_tokens": 400000,
"max_output_tokens": 128000,
"max_tokens": 128000,
"mode": "chat",
@ -31801,7 +31801,7 @@
"input_cost_per_token_flex": 1e-07,
"input_cost_per_token_batches": 1e-07,
"litellm_provider": "openai",
"max_input_tokens": 272000,
"max_input_tokens": 400000,
"max_output_tokens": 128000,
"max_tokens": 128000,
"mode": "chat",
@ -43488,6 +43488,7 @@
"supports_tool_choice": true
},
"together_ai/openai/gpt-oss-20b": {
"deprecation_date": "2026-09-14",
"input_cost_per_token": 5e-08,
"litellm_provider": "together_ai",
"max_input_tokens": 131072,
@ -43780,6 +43781,7 @@
"source": "https://docs.together.ai/docs/serverless-models"
},
"together_ai/google/gemma-4-31B-it": {
"deprecation_date": "2026-09-14",
"input_cost_per_token": 3.9e-07,
"litellm_provider": "together_ai",
"max_input_tokens": 262144,
@ -43794,6 +43796,7 @@
"supports_vision": true
},
"together_ai/intfloat/multilingual-e5-large-instruct": {
"deprecation_date": "2026-09-14",
"input_cost_per_token": 2e-08,
"litellm_provider": "together_ai",
"max_input_tokens": 514,
@ -43906,6 +43909,7 @@
"supports_tool_choice": true
},
"together_ai/thinkingmachines/Inkling-Small": {
"deprecation_date": "2026-09-14",
"cache_read_input_token_cost": 1e-07,
"input_cost_per_token": 5e-07,
"litellm_provider": "together_ai",
@ -60832,6 +60836,7 @@
"source": "https://docs.together.ai/docs/serverless-models"
},
"together_ai/moonshotai/Kimi-K2.6": {
"deprecation_date": "2026-08-19",
"input_cost_per_token": 1.2e-06,
"output_cost_per_token": 4.5e-06,
"cache_read_input_token_cost": 2e-07,
@ -60858,6 +60863,7 @@
"source": "https://api.together.xyz/v1/models"
},
"together_ai/zai-org/GLM-5": {
"deprecation_date": "2026-06-22",
"input_cost_per_token": 1e-06,
"output_cost_per_token": 3.2e-06,
"litellm_provider": "together_ai",
@ -60866,6 +60872,7 @@
"source": "https://api.together.xyz/v1/models"
},
"together_ai/zai-org/GLM-5.1": {
"deprecation_date": "2026-07-10",
"input_cost_per_token": 1.4e-06,
"output_cost_per_token": 4.4e-06,
"cache_read_input_token_cost": 2.6e-07,
@ -60883,6 +60890,7 @@
"source": "https://api.together.xyz/v1/models"
},
"together_ai/Qwen/Qwen3-Coder-Next-FP8": {
"deprecation_date": "2026-05-14",
"input_cost_per_token": 5e-07,
"output_cost_per_token": 1.2e-06,
"litellm_provider": "together_ai",
@ -60891,6 +60899,7 @@
"source": "https://api.together.xyz/v1/models"
},
"together_ai/Qwen/Qwen3-VL-32B-Instruct": {
"deprecation_date": "2026-02-25",
"input_cost_per_token": 5e-07,
"output_cost_per_token": 1.5e-06,
"litellm_provider": "together_ai",
@ -60899,6 +60908,7 @@
"source": "https://api.together.xyz/v1/models"
},
"together_ai/Qwen/Qwen3-VL-8B-Instruct": {
"deprecation_date": "2026-04-16",
"input_cost_per_token": 1.8e-07,
"output_cost_per_token": 6.8e-07,
"litellm_provider": "together_ai",
@ -60931,6 +60941,7 @@
"source": "https://api.together.xyz/v1/models"
},
"together_ai/Qwen/QwQ-32B": {
"deprecation_date": "2025-11-13",
"input_cost_per_token": 1.2e-06,
"output_cost_per_token": 1.2e-06,
"litellm_provider": "together_ai",

41
model_prices_and_context_window.json
View file

@ -8730,7 +8730,7 @@
"supports_function_calling": true
},
"azure/o1": {
"deprecation_date": "2026-10-21",
"deprecation_date": "2026-11-19",
"cache_read_input_token_cost": 7.5e-06,
"input_cost_per_token": 1.5e-05,
"litellm_provider": "azure",
@ -8748,7 +8748,7 @@
},
"azure/o1-2024-12-17": {
"cache_read_input_token_cost": 7.5e-06,
"deprecation_date": "2026-10-21",
"deprecation_date": "2026-11-19",
"input_cost_per_token": 1.5e-05,
"litellm_provider": "azure",
"max_input_tokens": 200000,
@ -8825,7 +8825,7 @@
"supports_vision": false
},
"azure/o3": {
"deprecation_date": "2026-10-21",
"deprecation_date": "2026-11-19",
"cache_read_input_token_cost": 5e-07,
"input_cost_per_token": 2e-06,
"litellm_provider": "azure",
@ -8855,7 +8855,7 @@
"supports_vision": true
},
"azure/o3-2025-04-16": {
"deprecation_date": "2026-10-21",
"deprecation_date": "2026-11-19",
"cache_read_input_token_cost": 5e-07,
"input_cost_per_token": 2e-06,
"litellm_provider": "azure",
@ -8886,7 +8886,7 @@
},
"azure/o3-deep-research": {
"cache_read_input_token_cost": 2.5e-06,
"deprecation_date": "2026-12-26",
"deprecation_date": "2026-11-19",
"input_cost_per_token": 1e-05,
"litellm_provider": "azure",
"max_input_tokens": 200000,
@ -8923,7 +8923,7 @@
"supports_web_search": true
},
"azure/o3-mini": {
"deprecation_date": "2026-10-01",
"deprecation_date": "2026-11-19",
"cache_read_input_token_cost": 5.5e-07,
"input_cost_per_token": 1.1e-06,
"litellm_provider": "azure",
@ -8940,7 +8940,7 @@
},
"azure/o3-mini-2025-01-31": {
"cache_read_input_token_cost": 5.5e-07,
"deprecation_date": "2026-10-01",
"deprecation_date": "2026-11-19",
"input_cost_per_token": 1.1e-06,
"litellm_provider": "azure",
"max_input_tokens": 200000,
@ -8954,7 +8954,7 @@
"supports_vision": false
},
"azure/o3-pro": {
"deprecation_date": "2026-12-17",
"deprecation_date": "2026-11-19",
"input_cost_per_token": 2e-05,
"input_cost_per_token_batches": 1e-05,
"litellm_provider": "azure",
@ -8985,7 +8985,7 @@
"supports_vision": true
},
"azure/o3-pro-2025-06-10": {
"deprecation_date": "2026-12-17",
"deprecation_date": "2026-11-19",
"input_cost_per_token": 2e-05,
"input_cost_per_token_batches": 1e-05,
"litellm_provider": "azure",
@ -9016,7 +9016,7 @@
"supports_vision": true
},
"azure/o4-mini": {
"deprecation_date": "2026-10-16",
"deprecation_date": "2026-11-19",
"cache_read_input_token_cost": 2.75e-07,
"input_cost_per_token": 1.1e-06,
"litellm_provider": "azure",
@ -9047,7 +9047,7 @@
},
"azure/o4-mini-2025-04-16": {
"cache_read_input_token_cost": 2.75e-07,
"deprecation_date": "2026-10-16",
"deprecation_date": "2026-11-19",
"input_cost_per_token": 1.1e-06,
"litellm_provider": "azure",
"max_input_tokens": 200000,
@ -31650,7 +31650,7 @@
"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": 400000,
"max_output_tokens": 128000,
"max_tokens": 128000,
"mode": "chat",
@ -31702,7 +31702,7 @@
"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": 400000,
"max_output_tokens": 128000,
"max_tokens": 128000,
"mode": "chat",
@ -31752,7 +31752,7 @@
"input_cost_per_token_flex": 1e-07,
"input_cost_per_token_batches": 1e-07,
"litellm_provider": "openai",
"max_input_tokens": 272000,
"max_input_tokens": 400000,
"max_output_tokens": 128000,
"max_tokens": 128000,
"mode": "chat",
@ -31801,7 +31801,7 @@
"input_cost_per_token_flex": 1e-07,
"input_cost_per_token_batches": 1e-07,
"litellm_provider": "openai",
"max_input_tokens": 272000,
"max_input_tokens": 400000,
"max_output_tokens": 128000,
"max_tokens": 128000,
"mode": "chat",
@ -43488,6 +43488,7 @@
"supports_tool_choice": true
},
"together_ai/openai/gpt-oss-20b": {
"deprecation_date": "2026-09-14",
"input_cost_per_token": 5e-08,
"litellm_provider": "together_ai",
"max_input_tokens": 131072,
@ -43780,6 +43781,7 @@
"source": "https://docs.together.ai/docs/serverless-models"
},
"together_ai/google/gemma-4-31B-it": {
"deprecation_date": "2026-09-14",
"input_cost_per_token": 3.9e-07,
"litellm_provider": "together_ai",
"max_input_tokens": 262144,
@ -43794,6 +43796,7 @@
"supports_vision": true
},
"together_ai/intfloat/multilingual-e5-large-instruct": {
"deprecation_date": "2026-09-14",
"input_cost_per_token": 2e-08,
"litellm_provider": "together_ai",
"max_input_tokens": 514,
@ -43906,6 +43909,7 @@
"supports_tool_choice": true
},
"together_ai/thinkingmachines/Inkling-Small": {
"deprecation_date": "2026-09-14",
"cache_read_input_token_cost": 1e-07,
"input_cost_per_token": 5e-07,
"litellm_provider": "together_ai",
@ -60832,6 +60836,7 @@
"source": "https://docs.together.ai/docs/serverless-models"
},
"together_ai/moonshotai/Kimi-K2.6": {
"deprecation_date": "2026-08-19",
"input_cost_per_token": 1.2e-06,
"output_cost_per_token": 4.5e-06,
"cache_read_input_token_cost": 2e-07,
@ -60858,6 +60863,7 @@
"source": "https://api.together.xyz/v1/models"
},
"together_ai/zai-org/GLM-5": {
"deprecation_date": "2026-06-22",
"input_cost_per_token": 1e-06,
"output_cost_per_token": 3.2e-06,
"litellm_provider": "together_ai",
@ -60866,6 +60872,7 @@
"source": "https://api.together.xyz/v1/models"
},
"together_ai/zai-org/GLM-5.1": {
"deprecation_date": "2026-07-10",
"input_cost_per_token": 1.4e-06,
"output_cost_per_token": 4.4e-06,
"cache_read_input_token_cost": 2.6e-07,
@ -60883,6 +60890,7 @@
"source": "https://api.together.xyz/v1/models"
},
"together_ai/Qwen/Qwen3-Coder-Next-FP8": {
"deprecation_date": "2026-05-14",
"input_cost_per_token": 5e-07,
"output_cost_per_token": 1.2e-06,
"litellm_provider": "together_ai",
@ -60891,6 +60899,7 @@
"source": "https://api.together.xyz/v1/models"
},
"together_ai/Qwen/Qwen3-VL-32B-Instruct": {
"deprecation_date": "2026-02-25",
"input_cost_per_token": 5e-07,
"output_cost_per_token": 1.5e-06,
"litellm_provider": "together_ai",
@ -60899,6 +60908,7 @@
"source": "https://api.together.xyz/v1/models"
},
"together_ai/Qwen/Qwen3-VL-8B-Instruct": {
"deprecation_date": "2026-04-16",
"input_cost_per_token": 1.8e-07,
"output_cost_per_token": 6.8e-07,
"litellm_provider": "together_ai",
@ -60931,6 +60941,7 @@
"source": "https://api.together.xyz/v1/models"
},
"together_ai/Qwen/QwQ-32B": {
"deprecation_date": "2025-11-13",
"input_cost_per_token": 1.2e-06,
"output_cost_per_token": 1.2e-06,
"litellm_provider": "together_ai",