From b4212b949b586a4a40d5b4bbc00029776282790f Mon Sep 17 00:00:00 2001
From: "berriai-litellm-provider-info-sync[bot]"
 <328147090+berriai-litellm-provider-info-sync[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 06:01:14 +0000
Subject: [PATCH 1/3] chore(prices): sync prices for 5 providers: 34 models, 1
 new, 19 deprecated [1 with gaps]

fireworks_ai/accounts/fireworks/routers/glm-5p3-fast:
azure_ai/FW-Kimi-K3: input_cost_per_token, output_cost_per_token, cache_read_input_token_cost
wandb/deepseek-ai/DeepSeek-R1-0528: deprecation_date
wandb/deepseek-ai/DeepSeek-V3-0324: deprecation_date
wandb/deepseek-ai/DeepSeek-V4-Flash: deprecation_date
wandb/deepseek-ai/DeepSeek-V4-Pro: deprecation_date
together_ai/deepseek-ai/DeepSeek-V4.1-Flash:
azure/eu/gpt-5.5-2026-04-24:
gemini-3.8-live: supports_response_schema
gemini-3.8-live-extended-thinking: supports_response_schema
azure/gpt-5.5-2026-04-24:
azure/gpt-5.6-luna-2026-07-09:
azure/gpt-5.6-sol-2026-07-09:
azure/gpt-5.6-terra-2026-07-09:
azure/gpt-6-astra-2026-09-03:
wandb/ibm-granite/granite-4.1-8b: deprecation_date
wandb/JetBrains/Mellum2-12B-A2.5B-Instruct: deprecation_date
wandb/meta-llama/Llama-3.1-70B-Instruct: deprecation_date
wandb/meta-llama/Llama-4-Scout-17B-16E-Instruct: deprecation_date
wandb/microsoft/Phi-4-mini-instruct: deprecation_date
wandb/MiniMaxAI/MiniMax-M2.5: deprecation_date
wandb/moonshotai/Kimi-K2-Instruct: deprecation_date
wandb/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B: input_cost_per_token, output_cost_per_token, cache_read_input_token_cost
wandb/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B: input_cost_per_token, output_cost_per_token, cache_read_input_token_cost
wandb/OpenPipe/Qwen3-14B-Instruct: deprecation_date
wandb/Qwen/Qwen3-235B-A22B-Instruct-2507: deprecation_date
wandb/Qwen/Qwen3-235B-A22B-Thinking-2507: deprecation_date
wandb/Qwen/Qwen3-30B-A3B-Instruct-2507: deprecation_date
wandb/Qwen/Qwen3-Coder-480B-A35B-Instruct: deprecation_date
wandb/Qwen/Qwen3.5-35B-A3B: deprecation_date
wandb/Qwen/Qwen3.6-27B: deprecation_date
azure/us/gpt-5.5-2026-04-24:
wandb/zai-org/GLM-4.5: deprecation_date
wandb/zai-org/GLM-5.3-Flash: input_cost_per_token, output_cost_per_token, cache_read_input_token_cost, supports_function_calling, supports_tool_choice, supports_response_schema, supports_prompt_caching, supports_reasoning
---
 ...odel_prices_and_context_window_backup.json | 74 ++++++++++++++-----
 model_prices_and_context_window.json          | 74 ++++++++++++++-----
 2 files changed, 108 insertions(+), 40 deletions(-)

diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json
index e87a3fec99b..914209e13ce 100644
--- a/litellm/model_prices_and_context_window_backup.json
+++ b/litellm/model_prices_and_context_window_backup.json
@@ -7605,7 +7605,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
-        "source": "https://learn.microsoft.com/en-us/azure/ai-foundry/openai/concepts/models",
+        "source": "https://prices.azure.com/api/retail/prices?$filter=serviceName%20eq%20'Foundry%20Models'%20and%20armRegionName%20eq%20'eastus'%20and%20priceType%20eq%20'Consumption'",
         "supported_endpoints": [
             "/v1/chat/completions",
             "/v1/batch",
@@ -7733,7 +7733,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
-        "source": "https://learn.microsoft.com/en-us/azure/ai-foundry/openai/concepts/models",
+        "source": "https://prices.azure.com/api/retail/prices?$filter=serviceName%20eq%20'Foundry%20Models'%20and%20armRegionName%20eq%20'eastus'%20and%20priceType%20eq%20'Consumption'",
         "supported_endpoints": [
             "/v1/chat/completions",
             "/v1/batch",
@@ -7887,7 +7887,7 @@
         "supports_none_reasoning_effort": true,
         "supports_xhigh_reasoning_effort": true,
         "supports_minimal_reasoning_effort": false,
-        "source": "https://learn.microsoft.com/en-us/azure/ai-foundry/openai/concepts/models"
+        "source": "https://prices.azure.com/api/retail/prices?$filter=serviceName%20eq%20'Foundry%20Models'%20and%20armRegionName%20eq%20'eastus'%20and%20priceType%20eq%20'Consumption'"
     },
     "azure/gpt-6-astra": {
         "cache_creation_input_token_cost": 1.25e-05,
@@ -7956,7 +7956,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
-        "source": "https://learn.microsoft.com/en-us/azure/ai-foundry/openai/concepts/models",
+        "source": "https://prices.azure.com/api/retail/prices?$filter=serviceName%20eq%20'Foundry%20Models'%20and%20armRegionName%20eq%20'eastus'%20and%20priceType%20eq%20'Consumption'",
         "supported_endpoints": [
             "/v1/chat/completions",
             "/v1/responses"
@@ -8856,7 +8856,7 @@
         "supports_none_reasoning_effort": true,
         "supports_xhigh_reasoning_effort": true,
         "supports_minimal_reasoning_effort": false,
-        "source": "https://learn.microsoft.com/en-us/azure/ai-foundry/openai/concepts/models"
+        "source": "https://prices.azure.com/api/retail/prices?$filter=serviceName%20eq%20'Foundry%20Models'%20and%20armRegionName%20eq%20'eastus'%20and%20priceType%20eq%20'Consumption'"
     },
     "azure/us/gpt-5.5-2026-04-23": {
         "cache_read_input_token_cost": 5.5e-07,
@@ -8955,7 +8955,7 @@
         "supports_none_reasoning_effort": true,
         "supports_xhigh_reasoning_effort": true,
         "supports_minimal_reasoning_effort": false,
-        "source": "https://learn.microsoft.com/en-us/azure/ai-foundry/openai/concepts/models"
+        "source": "https://prices.azure.com/api/retail/prices?$filter=serviceName%20eq%20'Foundry%20Models'%20and%20armRegionName%20eq%20'eastus'%20and%20priceType%20eq%20'Consumption'"
     },
     "azure/eu/gpt-5.5-2026-04-23": {
         "cache_read_input_token_cost": 5.5e-07,
@@ -9054,7 +9054,7 @@
         "supports_none_reasoning_effort": true,
         "supports_xhigh_reasoning_effort": true,
         "supports_minimal_reasoning_effort": false,
-        "source": "https://learn.microsoft.com/en-us/azure/ai-foundry/openai/concepts/models"
+        "source": "https://prices.azure.com/api/retail/prices?$filter=serviceName%20eq%20'Foundry%20Models'%20and%20armRegionName%20eq%20'eastus'%20and%20priceType%20eq%20'Consumption'"
     },
     "azure/gpt-5.5-pro": {
         "cache_read_input_token_cost": 3e-06,
@@ -10987,14 +10987,14 @@
         "supports_vision": true
     },
     "azure_ai/FW-Kimi-K3": {
-        "cache_read_input_token_cost": 3.3e-07,
-        "input_cost_per_token": 3.3e-06,
+        "cache_read_input_token_cost": 3e-07,
+        "input_cost_per_token": 3e-06,
         "litellm_provider": "azure_ai",
         "max_input_tokens": 1048576,
         "max_output_tokens": 131072,
         "max_tokens": 131072,
         "mode": "chat",
-        "output_cost_per_token": 1.65e-05,
+        "output_cost_per_token": 1.5e-05,
         "reasoning_effort_levels": [
             "low",
             "high",
@@ -45489,7 +45489,7 @@
         "max_tokens": 1048576,
         "mode": "chat",
         "output_cost_per_token": 1.2e-06,
-        "source": "https://api.together.xyz/v1/models",
+        "source": "https://api.together.ai/v1/models",
         "supports_function_calling": true,
         "supports_prompt_caching": true,
         "supports_response_schema": true,
@@ -50579,6 +50579,7 @@
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
     "wandb/zai-org/GLM-4.5": {
+        "deprecation_date": "2026-03-04",
         "supports_reasoning": true,
         "max_tokens": 131072,
         "max_input_tokens": 131072,
@@ -50589,6 +50590,7 @@
         "mode": "chat"
     },
     "wandb/Qwen/Qwen3-235B-A22B-Instruct-2507": {
+        "deprecation_date": "2026-08-04",
         "max_tokens": 262144,
         "max_input_tokens": 262144,
         "max_output_tokens": 262144,
@@ -50598,6 +50600,7 @@
         "mode": "chat"
     },
     "wandb/Qwen/Qwen3-Coder-480B-A35B-Instruct": {
+        "deprecation_date": "2026-08-25",
         "max_tokens": 262144,
         "max_input_tokens": 262144,
         "max_output_tokens": 262144,
@@ -50608,6 +50611,7 @@
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
     "wandb/Qwen/Qwen3-235B-A22B-Thinking-2507": {
+        "deprecation_date": "2026-08-04",
         "supports_reasoning": true,
         "max_tokens": 262144,
         "max_input_tokens": 262144,
@@ -50618,6 +50622,7 @@
         "mode": "chat"
     },
     "wandb/moonshotai/Kimi-K2-Instruct": {
+        "deprecation_date": "2026-03-04",
         "max_tokens": 128000,
         "max_input_tokens": 128000,
         "max_output_tokens": 128000,
@@ -50642,6 +50647,7 @@
         "supports_vision": true
     },
     "wandb/MiniMaxAI/MiniMax-M2.5": {
+        "deprecation_date": "2026-08-25",
         "max_tokens": 197000,
         "max_input_tokens": 197000,
         "max_output_tokens": 197000,
@@ -50676,6 +50682,7 @@
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
     "wandb/deepseek-ai/DeepSeek-R1-0528": {
+        "deprecation_date": "2026-03-04",
         "supports_reasoning": true,
         "max_tokens": 161000,
         "max_input_tokens": 161000,
@@ -50686,6 +50693,7 @@
         "mode": "chat"
     },
     "wandb/deepseek-ai/DeepSeek-V3-0324": {
+        "deprecation_date": "2026-03-04",
         "max_tokens": 161000,
         "max_input_tokens": 161000,
         "max_output_tokens": 161000,
@@ -50705,6 +50713,7 @@
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
     "wandb/meta-llama/Llama-4-Scout-17B-16E-Instruct": {
+        "deprecation_date": "2026-04-21",
         "max_tokens": 64000,
         "max_input_tokens": 64000,
         "max_output_tokens": 64000,
@@ -50714,6 +50723,7 @@
         "mode": "chat"
     },
     "wandb/microsoft/Phi-4-mini-instruct": {
+        "deprecation_date": "2026-08-04",
         "max_tokens": 128000,
         "max_input_tokens": 128000,
         "max_output_tokens": 128000,
@@ -56678,7 +56688,8 @@
         "supports_function_calling": true,
         "supports_vision": true,
         "supports_web_search": true,
-        "gemini_audio_only_live": true
+        "gemini_audio_only_live": true,
+        "supports_response_schema": false
     },
     "gemini-3.8-live-extended-thinking": {
         "input_cost_per_audio_token": 3e-06,
@@ -56712,7 +56723,8 @@
         "supports_vision": true,
         "supports_web_search": true,
         "gemini_audio_only_live": true,
-        "supports_reasoning": true
+        "supports_reasoning": true,
+        "supports_response_schema": false
     },
     "gemini/gemini-2.5-flash-native-audio-latest": {
         "input_cost_per_audio_token": 3e-06,
@@ -60960,6 +60972,7 @@
         "input_cost_per_token": 1.4e-07,
         "output_cost_per_token": 2.8e-07,
         "cache_read_input_token_cost": 7e-08,
+        "deprecation_date": "2026-10-05",
         "supports_prompt_caching": true,
         "litellm_provider": "wandb",
         "mode": "chat",
@@ -60986,6 +60999,7 @@
         "input_cost_per_token": 1.15e-06,
         "output_cost_per_token": 2.55e-06,
         "cache_read_input_token_cost": 2e-07,
+        "deprecation_date": "2026-10-05",
         "supports_prompt_caching": true,
         "litellm_provider": "wandb",
         "mode": "chat",
@@ -61004,6 +61018,7 @@
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
     "wandb/ibm-granite/granite-4.1-8b": {
+        "deprecation_date": "2026-10-05",
         "max_tokens": 131072,
         "max_input_tokens": 131072,
         "input_cost_per_token": 5e-08,
@@ -61014,6 +61029,7 @@
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
     "wandb/JetBrains/Mellum2-12B-A2.5B-Instruct": {
+        "deprecation_date": "2026-10-05",
         "max_tokens": 131072,
         "max_input_tokens": 131072,
         "input_cost_per_token": 5e-08,
@@ -61024,6 +61040,7 @@
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
     "wandb/meta-llama/Llama-3.1-70B-Instruct": {
+        "deprecation_date": "2026-10-05",
         "max_tokens": 128000,
         "max_input_tokens": 128000,
         "input_cost_per_token": 8e-07,
@@ -61076,9 +61093,9 @@
         "supports_reasoning": true,
         "max_tokens": 262144,
         "max_input_tokens": 262144,
-        "input_cost_per_token": 1e-07,
-        "output_cost_per_token": 2.5e-07,
-        "cache_read_input_token_cost": 5e-08,
+        "input_cost_per_token": 7e-08,
+        "output_cost_per_token": 2e-07,
+        "cache_read_input_token_cost": 4e-08,
         "supports_prompt_caching": true,
         "litellm_provider": "wandb",
         "mode": "chat",
@@ -61089,9 +61106,9 @@
         "supports_reasoning": true,
         "max_tokens": 262144,
         "max_input_tokens": 262144,
-        "input_cost_per_token": 7.5e-07,
-        "output_cost_per_token": 2.75e-06,
-        "cache_read_input_token_cost": 1.5e-07,
+        "input_cost_per_token": 5e-07,
+        "output_cost_per_token": 2.15e-06,
+        "cache_read_input_token_cost": 1e-07,
         "supports_prompt_caching": true,
         "litellm_provider": "wandb",
         "mode": "chat",
@@ -61099,6 +61116,7 @@
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
     "wandb/OpenPipe/Qwen3-14B-Instruct": {
+        "deprecation_date": "2026-10-05",
         "max_tokens": 32768,
         "max_input_tokens": 32768,
         "input_cost_per_token": 5e-08,
@@ -61139,6 +61157,7 @@
         "input_cost_per_token": 6e-07,
         "output_cost_per_token": 3.6e-06,
         "cache_read_input_token_cost": 1.2e-07,
+        "deprecation_date": "2026-10-05",
         "supports_prompt_caching": true,
         "litellm_provider": "wandb",
         "mode": "chat",
@@ -61146,6 +61165,7 @@
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
     "wandb/Qwen/Qwen3.5-35B-A3B": {
+        "deprecation_date": "2026-10-05",
         "supports_reasoning": true,
         "max_tokens": 262144,
         "max_input_tokens": 262144,
@@ -61157,6 +61177,7 @@
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
     "wandb/Qwen/Qwen3-30B-A3B-Instruct-2507": {
+        "deprecation_date": "2026-10-05",
         "max_tokens": 262144,
         "max_input_tokens": 262144,
         "input_cost_per_token": 1e-07,
@@ -62852,7 +62873,7 @@
         "max_tokens": 1048576,
         "mode": "chat",
         "output_cost_per_token": 6.6e-06,
-        "source": "https://docs.fireworks.ai/serverless/pricing",
+        "source": "https://api.fireworks.ai/v1/serverless/models",
         "supports_function_calling": true,
         "supports_reasoning": true,
         "supports_response_schema": true,
@@ -69144,5 +69165,18 @@
         "supports_prompt_caching": true,
         "supports_reasoning": true,
         "supports_vision": true
+    },
+    "wandb/zai-org/GLM-5.3-Flash": {
+        "cache_read_input_token_cost": 5e-08,
+        "input_cost_per_token": 1.5e-07,
+        "litellm_provider": "wandb",
+        "mode": "chat",
+        "output_cost_per_token": 5e-07,
+        "source": "https://wandb.ai/site/pricing/tokens/",
+        "supports_function_calling": true,
+        "supports_prompt_caching": true,
+        "supports_reasoning": true,
+        "supports_response_schema": true,
+        "supports_tool_choice": true
     }
 }
diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json
index e87a3fec99b..914209e13ce 100644
--- a/model_prices_and_context_window.json
+++ b/model_prices_and_context_window.json
@@ -7605,7 +7605,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
-        "source": "https://learn.microsoft.com/en-us/azure/ai-foundry/openai/concepts/models",
+        "source": "https://prices.azure.com/api/retail/prices?$filter=serviceName%20eq%20'Foundry%20Models'%20and%20armRegionName%20eq%20'eastus'%20and%20priceType%20eq%20'Consumption'",
         "supported_endpoints": [
             "/v1/chat/completions",
             "/v1/batch",
@@ -7733,7 +7733,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
-        "source": "https://learn.microsoft.com/en-us/azure/ai-foundry/openai/concepts/models",
+        "source": "https://prices.azure.com/api/retail/prices?$filter=serviceName%20eq%20'Foundry%20Models'%20and%20armRegionName%20eq%20'eastus'%20and%20priceType%20eq%20'Consumption'",
         "supported_endpoints": [
             "/v1/chat/completions",
             "/v1/batch",
@@ -7887,7 +7887,7 @@
         "supports_none_reasoning_effort": true,
         "supports_xhigh_reasoning_effort": true,
         "supports_minimal_reasoning_effort": false,
-        "source": "https://learn.microsoft.com/en-us/azure/ai-foundry/openai/concepts/models"
+        "source": "https://prices.azure.com/api/retail/prices?$filter=serviceName%20eq%20'Foundry%20Models'%20and%20armRegionName%20eq%20'eastus'%20and%20priceType%20eq%20'Consumption'"
     },
     "azure/gpt-6-astra": {
         "cache_creation_input_token_cost": 1.25e-05,
@@ -7956,7 +7956,7 @@
             "search_context_size_low": 0.01,
             "search_context_size_medium": 0.01
         },
-        "source": "https://learn.microsoft.com/en-us/azure/ai-foundry/openai/concepts/models",
+        "source": "https://prices.azure.com/api/retail/prices?$filter=serviceName%20eq%20'Foundry%20Models'%20and%20armRegionName%20eq%20'eastus'%20and%20priceType%20eq%20'Consumption'",
         "supported_endpoints": [
             "/v1/chat/completions",
             "/v1/responses"
@@ -8856,7 +8856,7 @@
         "supports_none_reasoning_effort": true,
         "supports_xhigh_reasoning_effort": true,
         "supports_minimal_reasoning_effort": false,
-        "source": "https://learn.microsoft.com/en-us/azure/ai-foundry/openai/concepts/models"
+        "source": "https://prices.azure.com/api/retail/prices?$filter=serviceName%20eq%20'Foundry%20Models'%20and%20armRegionName%20eq%20'eastus'%20and%20priceType%20eq%20'Consumption'"
     },
     "azure/us/gpt-5.5-2026-04-23": {
         "cache_read_input_token_cost": 5.5e-07,
@@ -8955,7 +8955,7 @@
         "supports_none_reasoning_effort": true,
         "supports_xhigh_reasoning_effort": true,
         "supports_minimal_reasoning_effort": false,
-        "source": "https://learn.microsoft.com/en-us/azure/ai-foundry/openai/concepts/models"
+        "source": "https://prices.azure.com/api/retail/prices?$filter=serviceName%20eq%20'Foundry%20Models'%20and%20armRegionName%20eq%20'eastus'%20and%20priceType%20eq%20'Consumption'"
     },
     "azure/eu/gpt-5.5-2026-04-23": {
         "cache_read_input_token_cost": 5.5e-07,
@@ -9054,7 +9054,7 @@
         "supports_none_reasoning_effort": true,
         "supports_xhigh_reasoning_effort": true,
         "supports_minimal_reasoning_effort": false,
-        "source": "https://learn.microsoft.com/en-us/azure/ai-foundry/openai/concepts/models"
+        "source": "https://prices.azure.com/api/retail/prices?$filter=serviceName%20eq%20'Foundry%20Models'%20and%20armRegionName%20eq%20'eastus'%20and%20priceType%20eq%20'Consumption'"
     },
     "azure/gpt-5.5-pro": {
         "cache_read_input_token_cost": 3e-06,
@@ -10987,14 +10987,14 @@
         "supports_vision": true
     },
     "azure_ai/FW-Kimi-K3": {
-        "cache_read_input_token_cost": 3.3e-07,
-        "input_cost_per_token": 3.3e-06,
+        "cache_read_input_token_cost": 3e-07,
+        "input_cost_per_token": 3e-06,
         "litellm_provider": "azure_ai",
         "max_input_tokens": 1048576,
         "max_output_tokens": 131072,
         "max_tokens": 131072,
         "mode": "chat",
-        "output_cost_per_token": 1.65e-05,
+        "output_cost_per_token": 1.5e-05,
         "reasoning_effort_levels": [
             "low",
             "high",
@@ -45489,7 +45489,7 @@
         "max_tokens": 1048576,
         "mode": "chat",
         "output_cost_per_token": 1.2e-06,
-        "source": "https://api.together.xyz/v1/models",
+        "source": "https://api.together.ai/v1/models",
         "supports_function_calling": true,
         "supports_prompt_caching": true,
         "supports_response_schema": true,
@@ -50579,6 +50579,7 @@
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
     "wandb/zai-org/GLM-4.5": {
+        "deprecation_date": "2026-03-04",
         "supports_reasoning": true,
         "max_tokens": 131072,
         "max_input_tokens": 131072,
@@ -50589,6 +50590,7 @@
         "mode": "chat"
     },
     "wandb/Qwen/Qwen3-235B-A22B-Instruct-2507": {
+        "deprecation_date": "2026-08-04",
         "max_tokens": 262144,
         "max_input_tokens": 262144,
         "max_output_tokens": 262144,
@@ -50598,6 +50600,7 @@
         "mode": "chat"
     },
     "wandb/Qwen/Qwen3-Coder-480B-A35B-Instruct": {
+        "deprecation_date": "2026-08-25",
         "max_tokens": 262144,
         "max_input_tokens": 262144,
         "max_output_tokens": 262144,
@@ -50608,6 +50611,7 @@
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
     "wandb/Qwen/Qwen3-235B-A22B-Thinking-2507": {
+        "deprecation_date": "2026-08-04",
         "supports_reasoning": true,
         "max_tokens": 262144,
         "max_input_tokens": 262144,
@@ -50618,6 +50622,7 @@
         "mode": "chat"
     },
     "wandb/moonshotai/Kimi-K2-Instruct": {
+        "deprecation_date": "2026-03-04",
         "max_tokens": 128000,
         "max_input_tokens": 128000,
         "max_output_tokens": 128000,
@@ -50642,6 +50647,7 @@
         "supports_vision": true
     },
     "wandb/MiniMaxAI/MiniMax-M2.5": {
+        "deprecation_date": "2026-08-25",
         "max_tokens": 197000,
         "max_input_tokens": 197000,
         "max_output_tokens": 197000,
@@ -50676,6 +50682,7 @@
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
     "wandb/deepseek-ai/DeepSeek-R1-0528": {
+        "deprecation_date": "2026-03-04",
         "supports_reasoning": true,
         "max_tokens": 161000,
         "max_input_tokens": 161000,
@@ -50686,6 +50693,7 @@
         "mode": "chat"
     },
     "wandb/deepseek-ai/DeepSeek-V3-0324": {
+        "deprecation_date": "2026-03-04",
         "max_tokens": 161000,
         "max_input_tokens": 161000,
         "max_output_tokens": 161000,
@@ -50705,6 +50713,7 @@
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
     "wandb/meta-llama/Llama-4-Scout-17B-16E-Instruct": {
+        "deprecation_date": "2026-04-21",
         "max_tokens": 64000,
         "max_input_tokens": 64000,
         "max_output_tokens": 64000,
@@ -50714,6 +50723,7 @@
         "mode": "chat"
     },
     "wandb/microsoft/Phi-4-mini-instruct": {
+        "deprecation_date": "2026-08-04",
         "max_tokens": 128000,
         "max_input_tokens": 128000,
         "max_output_tokens": 128000,
@@ -56678,7 +56688,8 @@
         "supports_function_calling": true,
         "supports_vision": true,
         "supports_web_search": true,
-        "gemini_audio_only_live": true
+        "gemini_audio_only_live": true,
+        "supports_response_schema": false
     },
     "gemini-3.8-live-extended-thinking": {
         "input_cost_per_audio_token": 3e-06,
@@ -56712,7 +56723,8 @@
         "supports_vision": true,
         "supports_web_search": true,
         "gemini_audio_only_live": true,
-        "supports_reasoning": true
+        "supports_reasoning": true,
+        "supports_response_schema": false
     },
     "gemini/gemini-2.5-flash-native-audio-latest": {
         "input_cost_per_audio_token": 3e-06,
@@ -60960,6 +60972,7 @@
         "input_cost_per_token": 1.4e-07,
         "output_cost_per_token": 2.8e-07,
         "cache_read_input_token_cost": 7e-08,
+        "deprecation_date": "2026-10-05",
         "supports_prompt_caching": true,
         "litellm_provider": "wandb",
         "mode": "chat",
@@ -60986,6 +60999,7 @@
         "input_cost_per_token": 1.15e-06,
         "output_cost_per_token": 2.55e-06,
         "cache_read_input_token_cost": 2e-07,
+        "deprecation_date": "2026-10-05",
         "supports_prompt_caching": true,
         "litellm_provider": "wandb",
         "mode": "chat",
@@ -61004,6 +61018,7 @@
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
     "wandb/ibm-granite/granite-4.1-8b": {
+        "deprecation_date": "2026-10-05",
         "max_tokens": 131072,
         "max_input_tokens": 131072,
         "input_cost_per_token": 5e-08,
@@ -61014,6 +61029,7 @@
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
     "wandb/JetBrains/Mellum2-12B-A2.5B-Instruct": {
+        "deprecation_date": "2026-10-05",
         "max_tokens": 131072,
         "max_input_tokens": 131072,
         "input_cost_per_token": 5e-08,
@@ -61024,6 +61040,7 @@
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
     "wandb/meta-llama/Llama-3.1-70B-Instruct": {
+        "deprecation_date": "2026-10-05",
         "max_tokens": 128000,
         "max_input_tokens": 128000,
         "input_cost_per_token": 8e-07,
@@ -61076,9 +61093,9 @@
         "supports_reasoning": true,
         "max_tokens": 262144,
         "max_input_tokens": 262144,
-        "input_cost_per_token": 1e-07,
-        "output_cost_per_token": 2.5e-07,
-        "cache_read_input_token_cost": 5e-08,
+        "input_cost_per_token": 7e-08,
+        "output_cost_per_token": 2e-07,
+        "cache_read_input_token_cost": 4e-08,
         "supports_prompt_caching": true,
         "litellm_provider": "wandb",
         "mode": "chat",
@@ -61089,9 +61106,9 @@
         "supports_reasoning": true,
         "max_tokens": 262144,
         "max_input_tokens": 262144,
-        "input_cost_per_token": 7.5e-07,
-        "output_cost_per_token": 2.75e-06,
-        "cache_read_input_token_cost": 1.5e-07,
+        "input_cost_per_token": 5e-07,
+        "output_cost_per_token": 2.15e-06,
+        "cache_read_input_token_cost": 1e-07,
         "supports_prompt_caching": true,
         "litellm_provider": "wandb",
         "mode": "chat",
@@ -61099,6 +61116,7 @@
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
     "wandb/OpenPipe/Qwen3-14B-Instruct": {
+        "deprecation_date": "2026-10-05",
         "max_tokens": 32768,
         "max_input_tokens": 32768,
         "input_cost_per_token": 5e-08,
@@ -61139,6 +61157,7 @@
         "input_cost_per_token": 6e-07,
         "output_cost_per_token": 3.6e-06,
         "cache_read_input_token_cost": 1.2e-07,
+        "deprecation_date": "2026-10-05",
         "supports_prompt_caching": true,
         "litellm_provider": "wandb",
         "mode": "chat",
@@ -61146,6 +61165,7 @@
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
     "wandb/Qwen/Qwen3.5-35B-A3B": {
+        "deprecation_date": "2026-10-05",
         "supports_reasoning": true,
         "max_tokens": 262144,
         "max_input_tokens": 262144,
@@ -61157,6 +61177,7 @@
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
     "wandb/Qwen/Qwen3-30B-A3B-Instruct-2507": {
+        "deprecation_date": "2026-10-05",
         "max_tokens": 262144,
         "max_input_tokens": 262144,
         "input_cost_per_token": 1e-07,
@@ -62852,7 +62873,7 @@
         "max_tokens": 1048576,
         "mode": "chat",
         "output_cost_per_token": 6.6e-06,
-        "source": "https://docs.fireworks.ai/serverless/pricing",
+        "source": "https://api.fireworks.ai/v1/serverless/models",
         "supports_function_calling": true,
         "supports_reasoning": true,
         "supports_response_schema": true,
@@ -69144,5 +69165,18 @@
         "supports_prompt_caching": true,
         "supports_reasoning": true,
         "supports_vision": true
+    },
+    "wandb/zai-org/GLM-5.3-Flash": {
+        "cache_read_input_token_cost": 5e-08,
+        "input_cost_per_token": 1.5e-07,
+        "litellm_provider": "wandb",
+        "mode": "chat",
+        "output_cost_per_token": 5e-07,
+        "source": "https://wandb.ai/site/pricing/tokens/",
+        "supports_function_calling": true,
+        "supports_prompt_caching": true,
+        "supports_reasoning": true,
+        "supports_response_schema": true,
+        "supports_tool_choice": true
     }
 }

From 5a82ed9bcabc0656e8c4055f5af9eb7b004ea4e6 Mon Sep 17 00:00:00 2001
From: "berriai-litellm-provider-info-sync[bot]"
 <328147090+berriai-litellm-provider-info-sync[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 06:31:05 +0000
Subject: [PATCH 2/3] chore(prices): sync prices for 2 providers: 27 models

fireworks_ai/accounts/fireworks/models/minimax-m3: supports_vision
fireworks_ai/minimax-m3: supports_vision
wandb/deepseek-ai/DeepSeek-V4-Flash: max_input_tokens
wandb/deepseek-ai/DeepSeek-V4-Flash-0731: max_input_tokens
wandb/deepseek-ai/DeepSeek-V4-Pro: max_input_tokens
wandb/deepseek-ai/DeepSeek-V4-Pro-0813: max_input_tokens
wandb/google/gemma-4-31B-it: max_input_tokens
wandb/ibm-granite/granite-4.1-8b: max_input_tokens
wandb/ibm-granite/granite-4.2-8b: max_input_tokens
wandb/JetBrains/Mellum2-12B-A2.5B-Instruct: max_input_tokens
wandb/meta-llama/Llama-3.1-70B-Instruct: max_input_tokens
wandb/meta-llama/Llama-3.1-8B-Instruct: max_input_tokens
wandb/MiniMaxAI/MiniMax-M3: max_input_tokens
wandb/moonshotai/Kimi-K2.6: max_input_tokens
wandb/moonshotai/Kimi-K2.7-Code: max_input_tokens
wandb/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B: max_input_tokens
wandb/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B: max_input_tokens
wandb/openai/gpt-oss-120b: max_input_tokens
wandb/openai/gpt-oss-20b: max_input_tokens
wandb/OpenPipe/Qwen3-14B-Instruct: max_input_tokens
wandb/Qwen/Qwen3-30B-A3B-Instruct-2507: max_input_tokens
wandb/Qwen/Qwen3.5-35B-A3B: max_input_tokens
wandb/Qwen/Qwen3.6-27B: max_input_tokens
wandb/Qwen/Qwen3.6-35B-A3B: max_input_tokens
wandb/Qwen/Qwen3.8-27B: max_input_tokens
wandb/zai-org/GLM-5.2: max_input_tokens
wandb/zai-org/GLM-5.3-Flash: max_input_tokens
---
 ...odel_prices_and_context_window_backup.json | 51 ++++++++++---------
 model_prices_and_context_window.json          | 51 ++++++++++---------
 2 files changed, 54 insertions(+), 48 deletions(-)

diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json
index 914209e13ce..12ce1465123 100644
--- a/litellm/model_prices_and_context_window_backup.json
+++ b/litellm/model_prices_and_context_window_backup.json
@@ -23788,7 +23788,7 @@
         "supports_reasoning": true,
         "supports_response_schema": true,
         "supports_tool_choice": true,
-        "supports_vision": true
+        "supports_vision": false
     },
     "fireworks_ai/accounts/fireworks/models/mixtral-8x22b-instruct-hf": {
         "input_cost_per_token": 1.2e-06,
@@ -24114,7 +24114,7 @@
         "supports_reasoning": true,
         "supports_response_schema": true,
         "supports_tool_choice": true,
-        "supports_vision": true
+        "supports_vision": false
     },
     "fireworks_ai/qwen3p7-plus": {
         "cache_read_input_token_cost": 8e-08,
@@ -50559,7 +50559,7 @@
     "wandb/openai/gpt-oss-120b": {
         "supports_reasoning": true,
         "max_tokens": 131072,
-        "max_input_tokens": 131072,
+        "max_input_tokens": 131000,
         "max_output_tokens": 131072,
         "input_cost_per_token": 3e-08,
         "output_cost_per_token": 1.7e-07,
@@ -50570,7 +50570,7 @@
     "wandb/openai/gpt-oss-20b": {
         "supports_reasoning": true,
         "max_tokens": 131072,
-        "max_input_tokens": 131072,
+        "max_input_tokens": 131000,
         "max_output_tokens": 131072,
         "input_cost_per_token": 3e-08,
         "output_cost_per_token": 1.3e-07,
@@ -50662,7 +50662,7 @@
     },
     "wandb/meta-llama/Llama-3.1-8B-Instruct": {
         "max_tokens": 128000,
-        "max_input_tokens": 128000,
+        "max_input_tokens": 131000,
         "max_output_tokens": 128000,
         "input_cost_per_token": 2.2e-07,
         "output_cost_per_token": 2.2e-07,
@@ -60968,7 +60968,7 @@
     "wandb/deepseek-ai/DeepSeek-V4-Flash": {
         "supports_reasoning": true,
         "max_tokens": 1048576,
-        "max_input_tokens": 1048576,
+        "max_input_tokens": 1049000,
         "input_cost_per_token": 1.4e-07,
         "output_cost_per_token": 2.8e-07,
         "cache_read_input_token_cost": 7e-08,
@@ -60982,7 +60982,7 @@
     "wandb/deepseek-ai/DeepSeek-V4-Flash-0731": {
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 1.3e-07,
         "output_cost_per_token": 2.8e-07,
         "cache_read_input_token_cost": 7e-08,
@@ -60995,7 +60995,7 @@
     "wandb/deepseek-ai/DeepSeek-V4-Pro": {
         "supports_reasoning": true,
         "max_tokens": 1048576,
-        "max_input_tokens": 1048576,
+        "max_input_tokens": 1049000,
         "input_cost_per_token": 1.15e-06,
         "output_cost_per_token": 2.55e-06,
         "cache_read_input_token_cost": 2e-07,
@@ -61009,7 +61009,7 @@
     "wandb/google/gemma-4-31B-it": {
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 1e-07,
         "output_cost_per_token": 3.4e-07,
         "litellm_provider": "wandb",
@@ -61020,7 +61020,7 @@
     "wandb/ibm-granite/granite-4.1-8b": {
         "deprecation_date": "2026-10-05",
         "max_tokens": 131072,
-        "max_input_tokens": 131072,
+        "max_input_tokens": 131000,
         "input_cost_per_token": 5e-08,
         "output_cost_per_token": 1e-07,
         "litellm_provider": "wandb",
@@ -61031,7 +61031,7 @@
     "wandb/JetBrains/Mellum2-12B-A2.5B-Instruct": {
         "deprecation_date": "2026-10-05",
         "max_tokens": 131072,
-        "max_input_tokens": 131072,
+        "max_input_tokens": 131000,
         "input_cost_per_token": 5e-08,
         "output_cost_per_token": 1e-07,
         "litellm_provider": "wandb",
@@ -61042,7 +61042,7 @@
     "wandb/meta-llama/Llama-3.1-70B-Instruct": {
         "deprecation_date": "2026-10-05",
         "max_tokens": 128000,
-        "max_input_tokens": 128000,
+        "max_input_tokens": 131000,
         "input_cost_per_token": 8e-07,
         "output_cost_per_token": 8e-07,
         "litellm_provider": "wandb",
@@ -61053,7 +61053,7 @@
     "wandb/MiniMaxAI/MiniMax-M3": {
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 2.3e-07,
         "output_cost_per_token": 9.6e-07,
         "cache_read_input_token_cost": 5e-08,
@@ -61066,7 +61066,7 @@
     "wandb/moonshotai/Kimi-K2.7-Code": {
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 7.1e-07,
         "output_cost_per_token": 3.5e-06,
         "cache_read_input_token_cost": 1.5e-07,
@@ -61079,7 +61079,7 @@
     "wandb/moonshotai/Kimi-K2.6": {
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 6.5e-07,
         "output_cost_per_token": 3.41e-06,
         "cache_read_input_token_cost": 1.5e-07,
@@ -61092,7 +61092,7 @@
     "wandb/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B": {
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 7e-08,
         "output_cost_per_token": 2e-07,
         "cache_read_input_token_cost": 4e-08,
@@ -61105,7 +61105,7 @@
     "wandb/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B": {
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 5e-07,
         "output_cost_per_token": 2.15e-06,
         "cache_read_input_token_cost": 1e-07,
@@ -61118,7 +61118,7 @@
     "wandb/OpenPipe/Qwen3-14B-Instruct": {
         "deprecation_date": "2026-10-05",
         "max_tokens": 32768,
-        "max_input_tokens": 32768,
+        "max_input_tokens": 32800,
         "input_cost_per_token": 5e-08,
         "output_cost_per_token": 2.2e-07,
         "litellm_provider": "wandb",
@@ -61129,7 +61129,7 @@
     "wandb/Qwen/Qwen3.8-27B": {
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 4e-07,
         "output_cost_per_token": 3e-06,
         "cache_read_input_token_cost": 1.5e-07,
@@ -61142,7 +61142,7 @@
     "wandb/Qwen/Qwen3.6-35B-A3B": {
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 2.5e-07,
         "output_cost_per_token": 1.25e-06,
         "litellm_provider": "wandb",
@@ -61153,7 +61153,7 @@
     "wandb/Qwen/Qwen3.6-27B": {
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 6e-07,
         "output_cost_per_token": 3.6e-06,
         "cache_read_input_token_cost": 1.2e-07,
@@ -61168,7 +61168,7 @@
         "deprecation_date": "2026-10-05",
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 2.5e-07,
         "output_cost_per_token": 1.25e-06,
         "litellm_provider": "wandb",
@@ -61179,7 +61179,7 @@
     "wandb/Qwen/Qwen3-30B-A3B-Instruct-2507": {
         "deprecation_date": "2026-10-05",
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 1e-07,
         "output_cost_per_token": 3e-07,
         "litellm_provider": "wandb",
@@ -61194,6 +61194,7 @@
         "input_cost_per_token": 1.31e-06,
         "output_cost_per_token": 3.96e-06,
         "cache_read_input_token_cost": 4.4e-08,
+        "max_input_tokens": 1049000,
         "supports_prompt_caching": true,
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
@@ -61204,13 +61205,14 @@
         "input_cost_per_token": 1e-07,
         "output_cost_per_token": 1.5e-07,
         "cache_read_input_token_cost": 5e-08,
+        "max_input_tokens": 131000,
         "supports_prompt_caching": true,
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
     "wandb/zai-org/GLM-5.2": {
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 1049000,
         "input_cost_per_token": 7.6e-07,
         "output_cost_per_token": 2.42e-06,
         "cache_read_input_token_cost": 1.4e-07,
@@ -69170,6 +69172,7 @@
         "cache_read_input_token_cost": 5e-08,
         "input_cost_per_token": 1.5e-07,
         "litellm_provider": "wandb",
+        "max_input_tokens": 1049000,
         "mode": "chat",
         "output_cost_per_token": 5e-07,
         "source": "https://wandb.ai/site/pricing/tokens/",
diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json
index 914209e13ce..12ce1465123 100644
--- a/model_prices_and_context_window.json
+++ b/model_prices_and_context_window.json
@@ -23788,7 +23788,7 @@
         "supports_reasoning": true,
         "supports_response_schema": true,
         "supports_tool_choice": true,
-        "supports_vision": true
+        "supports_vision": false
     },
     "fireworks_ai/accounts/fireworks/models/mixtral-8x22b-instruct-hf": {
         "input_cost_per_token": 1.2e-06,
@@ -24114,7 +24114,7 @@
         "supports_reasoning": true,
         "supports_response_schema": true,
         "supports_tool_choice": true,
-        "supports_vision": true
+        "supports_vision": false
     },
     "fireworks_ai/qwen3p7-plus": {
         "cache_read_input_token_cost": 8e-08,
@@ -50559,7 +50559,7 @@
     "wandb/openai/gpt-oss-120b": {
         "supports_reasoning": true,
         "max_tokens": 131072,
-        "max_input_tokens": 131072,
+        "max_input_tokens": 131000,
         "max_output_tokens": 131072,
         "input_cost_per_token": 3e-08,
         "output_cost_per_token": 1.7e-07,
@@ -50570,7 +50570,7 @@
     "wandb/openai/gpt-oss-20b": {
         "supports_reasoning": true,
         "max_tokens": 131072,
-        "max_input_tokens": 131072,
+        "max_input_tokens": 131000,
         "max_output_tokens": 131072,
         "input_cost_per_token": 3e-08,
         "output_cost_per_token": 1.3e-07,
@@ -50662,7 +50662,7 @@
     },
     "wandb/meta-llama/Llama-3.1-8B-Instruct": {
         "max_tokens": 128000,
-        "max_input_tokens": 128000,
+        "max_input_tokens": 131000,
         "max_output_tokens": 128000,
         "input_cost_per_token": 2.2e-07,
         "output_cost_per_token": 2.2e-07,
@@ -60968,7 +60968,7 @@
     "wandb/deepseek-ai/DeepSeek-V4-Flash": {
         "supports_reasoning": true,
         "max_tokens": 1048576,
-        "max_input_tokens": 1048576,
+        "max_input_tokens": 1049000,
         "input_cost_per_token": 1.4e-07,
         "output_cost_per_token": 2.8e-07,
         "cache_read_input_token_cost": 7e-08,
@@ -60982,7 +60982,7 @@
     "wandb/deepseek-ai/DeepSeek-V4-Flash-0731": {
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 1.3e-07,
         "output_cost_per_token": 2.8e-07,
         "cache_read_input_token_cost": 7e-08,
@@ -60995,7 +60995,7 @@
     "wandb/deepseek-ai/DeepSeek-V4-Pro": {
         "supports_reasoning": true,
         "max_tokens": 1048576,
-        "max_input_tokens": 1048576,
+        "max_input_tokens": 1049000,
         "input_cost_per_token": 1.15e-06,
         "output_cost_per_token": 2.55e-06,
         "cache_read_input_token_cost": 2e-07,
@@ -61009,7 +61009,7 @@
     "wandb/google/gemma-4-31B-it": {
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 1e-07,
         "output_cost_per_token": 3.4e-07,
         "litellm_provider": "wandb",
@@ -61020,7 +61020,7 @@
     "wandb/ibm-granite/granite-4.1-8b": {
         "deprecation_date": "2026-10-05",
         "max_tokens": 131072,
-        "max_input_tokens": 131072,
+        "max_input_tokens": 131000,
         "input_cost_per_token": 5e-08,
         "output_cost_per_token": 1e-07,
         "litellm_provider": "wandb",
@@ -61031,7 +61031,7 @@
     "wandb/JetBrains/Mellum2-12B-A2.5B-Instruct": {
         "deprecation_date": "2026-10-05",
         "max_tokens": 131072,
-        "max_input_tokens": 131072,
+        "max_input_tokens": 131000,
         "input_cost_per_token": 5e-08,
         "output_cost_per_token": 1e-07,
         "litellm_provider": "wandb",
@@ -61042,7 +61042,7 @@
     "wandb/meta-llama/Llama-3.1-70B-Instruct": {
         "deprecation_date": "2026-10-05",
         "max_tokens": 128000,
-        "max_input_tokens": 128000,
+        "max_input_tokens": 131000,
         "input_cost_per_token": 8e-07,
         "output_cost_per_token": 8e-07,
         "litellm_provider": "wandb",
@@ -61053,7 +61053,7 @@
     "wandb/MiniMaxAI/MiniMax-M3": {
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 2.3e-07,
         "output_cost_per_token": 9.6e-07,
         "cache_read_input_token_cost": 5e-08,
@@ -61066,7 +61066,7 @@
     "wandb/moonshotai/Kimi-K2.7-Code": {
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 7.1e-07,
         "output_cost_per_token": 3.5e-06,
         "cache_read_input_token_cost": 1.5e-07,
@@ -61079,7 +61079,7 @@
     "wandb/moonshotai/Kimi-K2.6": {
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 6.5e-07,
         "output_cost_per_token": 3.41e-06,
         "cache_read_input_token_cost": 1.5e-07,
@@ -61092,7 +61092,7 @@
     "wandb/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B": {
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 7e-08,
         "output_cost_per_token": 2e-07,
         "cache_read_input_token_cost": 4e-08,
@@ -61105,7 +61105,7 @@
     "wandb/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B": {
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 5e-07,
         "output_cost_per_token": 2.15e-06,
         "cache_read_input_token_cost": 1e-07,
@@ -61118,7 +61118,7 @@
     "wandb/OpenPipe/Qwen3-14B-Instruct": {
         "deprecation_date": "2026-10-05",
         "max_tokens": 32768,
-        "max_input_tokens": 32768,
+        "max_input_tokens": 32800,
         "input_cost_per_token": 5e-08,
         "output_cost_per_token": 2.2e-07,
         "litellm_provider": "wandb",
@@ -61129,7 +61129,7 @@
     "wandb/Qwen/Qwen3.8-27B": {
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 4e-07,
         "output_cost_per_token": 3e-06,
         "cache_read_input_token_cost": 1.5e-07,
@@ -61142,7 +61142,7 @@
     "wandb/Qwen/Qwen3.6-35B-A3B": {
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 2.5e-07,
         "output_cost_per_token": 1.25e-06,
         "litellm_provider": "wandb",
@@ -61153,7 +61153,7 @@
     "wandb/Qwen/Qwen3.6-27B": {
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 6e-07,
         "output_cost_per_token": 3.6e-06,
         "cache_read_input_token_cost": 1.2e-07,
@@ -61168,7 +61168,7 @@
         "deprecation_date": "2026-10-05",
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 2.5e-07,
         "output_cost_per_token": 1.25e-06,
         "litellm_provider": "wandb",
@@ -61179,7 +61179,7 @@
     "wandb/Qwen/Qwen3-30B-A3B-Instruct-2507": {
         "deprecation_date": "2026-10-05",
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 262000,
         "input_cost_per_token": 1e-07,
         "output_cost_per_token": 3e-07,
         "litellm_provider": "wandb",
@@ -61194,6 +61194,7 @@
         "input_cost_per_token": 1.31e-06,
         "output_cost_per_token": 3.96e-06,
         "cache_read_input_token_cost": 4.4e-08,
+        "max_input_tokens": 1049000,
         "supports_prompt_caching": true,
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
@@ -61204,13 +61205,14 @@
         "input_cost_per_token": 1e-07,
         "output_cost_per_token": 1.5e-07,
         "cache_read_input_token_cost": 5e-08,
+        "max_input_tokens": 131000,
         "supports_prompt_caching": true,
         "source": "https://wandb.ai/site/pricing/tokens/"
     },
     "wandb/zai-org/GLM-5.2": {
         "supports_reasoning": true,
         "max_tokens": 262144,
-        "max_input_tokens": 262144,
+        "max_input_tokens": 1049000,
         "input_cost_per_token": 7.6e-07,
         "output_cost_per_token": 2.42e-06,
         "cache_read_input_token_cost": 1.4e-07,
@@ -69170,6 +69172,7 @@
         "cache_read_input_token_cost": 5e-08,
         "input_cost_per_token": 1.5e-07,
         "litellm_provider": "wandb",
+        "max_input_tokens": 1049000,
         "mode": "chat",
         "output_cost_per_token": 5e-07,
         "source": "https://wandb.ai/site/pricing/tokens/",

From 730195c6030921f0a3bf15ca4c95f2eb468a370c Mon Sep 17 00:00:00 2001
From: "berriai-litellm-provider-info-sync[bot]"
 <328147090+berriai-litellm-provider-info-sync[bot]@users.noreply.github.com>
Date: Thu, 17 Sep 2026 18:01:11 +0000
Subject: [PATCH 3/3] chore(prices): sync Together AI prices: 6 models, 6
 deprecated [sync failed: Google Gemini]

together_ai/deepseek-ai/DeepSeek-V4-Flash-0731: deprecation_date
together_ai/deepseek-ai/DeepSeek-V4-Pro-0813: deprecation_date
together_ai/google/gemma-4-31B-it: deprecation_date
together_ai/intfloat/multilingual-e5-large-instruct: deprecation_date
together_ai/openai/gpt-oss-20b: deprecation_date
together_ai/thinkingmachines/Inkling-Small: deprecation_date
---
 litellm/model_prices_and_context_window_backup.json | 10 ++++++----
 model_prices_and_context_window.json                | 10 ++++++----
 2 files changed, 12 insertions(+), 8 deletions(-)

diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json
index 12ce1465123..7212c3950d5 100644
--- a/litellm/model_prices_and_context_window_backup.json
+++ b/litellm/model_prices_and_context_window_backup.json
@@ -45231,7 +45231,7 @@
         "supports_tool_choice": true
     },
     "together_ai/openai/gpt-oss-20b": {
-        "deprecation_date": "2026-09-15",
+        "deprecation_date": "2026-09-14",
         "input_cost_per_token": 5e-08,
         "litellm_provider": "together_ai",
         "max_input_tokens": 131072,
@@ -45468,6 +45468,7 @@
     },
     "together_ai/deepseek-ai/DeepSeek-V4-Flash-0731": {
         "cache_read_input_token_cost": 3e-08,
+        "deprecation_date": "2026-09-29",
         "input_cost_per_token": 1.4e-07,
         "litellm_provider": "together_ai",
         "max_input_tokens": 1048576,
@@ -45514,6 +45515,7 @@
     },
     "together_ai/deepseek-ai/DeepSeek-V4-Pro-0813": {
         "cache_read_input_token_cost": 1.3e-07,
+        "deprecation_date": "2026-09-29",
         "input_cost_per_token": 1.32e-06,
         "litellm_provider": "together_ai",
         "max_input_tokens": 1048576,
@@ -45538,7 +45540,7 @@
         "source": "https://docs.together.ai/docs/serverless-models"
     },
     "together_ai/google/gemma-4-31B-it": {
-        "deprecation_date": "2026-09-15",
+        "deprecation_date": "2026-09-14",
         "input_cost_per_token": 3.9e-07,
         "litellm_provider": "together_ai",
         "max_input_tokens": 262144,
@@ -45553,7 +45555,7 @@
         "supports_vision": true
     },
     "together_ai/intfloat/multilingual-e5-large-instruct": {
-        "deprecation_date": "2026-09-15",
+        "deprecation_date": "2026-09-14",
         "input_cost_per_token": 2e-08,
         "litellm_provider": "together_ai",
         "max_input_tokens": 514,
@@ -45666,7 +45668,7 @@
         "supports_tool_choice": true
     },
     "together_ai/thinkingmachines/Inkling-Small": {
-        "deprecation_date": "2026-09-15",
+        "deprecation_date": "2026-09-14",
         "cache_read_input_token_cost": 1e-07,
         "input_cost_per_token": 5e-07,
         "litellm_provider": "together_ai",
diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json
index 12ce1465123..7212c3950d5 100644
--- a/model_prices_and_context_window.json
+++ b/model_prices_and_context_window.json
@@ -45231,7 +45231,7 @@
         "supports_tool_choice": true
     },
     "together_ai/openai/gpt-oss-20b": {
-        "deprecation_date": "2026-09-15",
+        "deprecation_date": "2026-09-14",
         "input_cost_per_token": 5e-08,
         "litellm_provider": "together_ai",
         "max_input_tokens": 131072,
@@ -45468,6 +45468,7 @@
     },
     "together_ai/deepseek-ai/DeepSeek-V4-Flash-0731": {
         "cache_read_input_token_cost": 3e-08,
+        "deprecation_date": "2026-09-29",
         "input_cost_per_token": 1.4e-07,
         "litellm_provider": "together_ai",
         "max_input_tokens": 1048576,
@@ -45514,6 +45515,7 @@
     },
     "together_ai/deepseek-ai/DeepSeek-V4-Pro-0813": {
         "cache_read_input_token_cost": 1.3e-07,
+        "deprecation_date": "2026-09-29",
         "input_cost_per_token": 1.32e-06,
         "litellm_provider": "together_ai",
         "max_input_tokens": 1048576,
@@ -45538,7 +45540,7 @@
         "source": "https://docs.together.ai/docs/serverless-models"
     },
     "together_ai/google/gemma-4-31B-it": {
-        "deprecation_date": "2026-09-15",
+        "deprecation_date": "2026-09-14",
         "input_cost_per_token": 3.9e-07,
         "litellm_provider": "together_ai",
         "max_input_tokens": 262144,
@@ -45553,7 +45555,7 @@
         "supports_vision": true
     },
     "together_ai/intfloat/multilingual-e5-large-instruct": {
-        "deprecation_date": "2026-09-15",
+        "deprecation_date": "2026-09-14",
         "input_cost_per_token": 2e-08,
         "litellm_provider": "together_ai",
         "max_input_tokens": 514,
@@ -45666,7 +45668,7 @@
         "supports_tool_choice": true
     },
     "together_ai/thinkingmachines/Inkling-Small": {
-        "deprecation_date": "2026-09-15",
+        "deprecation_date": "2026-09-14",
         "cache_read_input_token_cost": 1e-07,
         "input_cost_per_token": 5e-07,
         "litellm_provider": "together_ai",