diff --git a/packages/proxy/schema/index.ts b/packages/proxy/schema/index.ts index 01bad129..906f5265 100644 --- a/packages/proxy/schema/index.ts +++ b/packages/proxy/schema/index.ts @@ -364,7 +364,7 @@ export const AvailableEndpointTypes: { [name: string]: ModelEndpointType[] } = { "pixtral-large-latest": ["mistral"], "mistral-medium-latest": ["mistral"], "mistral-medium-2505": ["mistral"], - "pixtral-large-2411": ["mistral", "vertex"], + "pixtral-large-2411": ["mistral"], "mistral-small-latest": ["mistral"], "mistral-small-2501": ["mistral"], "codestral-2501": ["mistral"], @@ -947,7 +947,7 @@ export const AvailableEndpointTypes: { [name: string]: ModelEndpointType[] } = { "grok-code-fast-1": ["xAI"], "grok-code-fast": ["xAI"], "grok-code-fast-1-0825": ["xAI"], - "gemini-3.5-flash": ["google"], + "gemini-3.5-flash": ["google", "vertex"], "gemini-3.1-flash-image-preview": ["google"], "gemini-3.1-flash-lite": ["google"], "gemini-3.1-flash-lite-preview": ["google"], @@ -1043,7 +1043,7 @@ export const AvailableEndpointTypes: { [name: string]: ModelEndpointType[] } = { "databricks-meta-llama-3-3-70b-instruct": ["databricks"], "databricks-meta-llama-3-1-405b-instruct": ["databricks"], "databricks-meta-llama-3-1-8b-instruct": ["databricks"], - "openai/gpt-oss-120b": ["together", "groq", "baseten"], + "openai/gpt-oss-120b": ["groq", "together", "baseten"], "openai/gpt-oss-20b": ["groq"], // NOTE: We use groq pricing for this and Together pricing for the 120B model "o4-mini-deep-research-2025-06-26": ["openai", "azure"], "o4-mini-deep-research": ["openai", "azure"], @@ -1060,6 +1060,8 @@ export const AvailableEndpointTypes: { [name: string]: ModelEndpointType[] } = { "accounts/fireworks/models/deepseek-v2-lite-chat": ["fireworks"], "accounts/fireworks/models/kimi-k2-instruct": ["fireworks"], "grok-latest": ["xAI"], + "gemma-4-31b": ["cerebras"], + "gpt-5.4-mini-2026-03-17": ["openai", "azure"], }; const modelEndpointAvailableModels = getAvailableModels(); diff --git a/packages/proxy/schema/model_list.json b/packages/proxy/schema/model_list.json index 34173804..8b07db8f 100644 --- a/packages/proxy/schema/model_list.json +++ b/packages/proxy/schema/model_list.json @@ -78,7 +78,7 @@ "displayName": "GPT-5.4 mini (2026-03-17)", "reasoning": true, "parent": "gpt-5.4-mini", - "max_input_tokens": 272000, + "max_input_tokens": 400000, "max_output_tokens": 128000, "available_providers": [ "openai", @@ -164,7 +164,7 @@ "input_cache_read_cost_per_mil_tokens": 0.075, "displayName": "GPT-5.4 mini", "reasoning": true, - "max_input_tokens": 272000, + "max_input_tokens": 400000, "max_output_tokens": 128000, "available_providers": [ "openai", @@ -180,7 +180,7 @@ "input_cache_read_cost_per_mil_tokens": 0.02, "displayName": "GPT-5.4 nano", "reasoning": true, - "max_input_tokens": 272000, + "max_input_tokens": 400000, "max_output_tokens": 128000, "available_providers": [ "openai", @@ -1313,8 +1313,8 @@ "openai/gpt-oss-120b": { "format": "openai", "flavor": "chat", - "input_cost_per_mil_tokens": 0.15, - "output_cost_per_mil_tokens": 0.6, + "input_cost_per_mil_tokens": 0.1, + "output_cost_per_mil_tokens": 0.5, "input_cache_read_cost_per_mil_tokens": 0.1, "displayName": "OpenAI GPT-OSS (120B)", "reasoning": true, @@ -1655,7 +1655,7 @@ "us.anthropic.claude-sonnet-4-6" ], "max_input_tokens": 1000000, - "max_output_tokens": 64000, + "max_output_tokens": 128000, "available_providers": [ "anthropic" ] @@ -3249,6 +3249,7 @@ "displayName": "Qwen3-32B", "reasoning": true, "experimental": true, + "deprecation_date": "2026-07-17", "max_input_tokens": 131072, "max_output_tokens": 40960, "available_providers": [ @@ -4084,7 +4085,7 @@ "flavor": "chat", "input_cost_per_mil_tokens": 0.6, "output_cost_per_mil_tokens": 3.6, - "input_cache_read_cost_per_mil_tokens": 0.06, + "input_cache_read_cost_per_mil_tokens": 0.35, "displayName": "Qwen3.5 397B A17B", "max_input_tokens": 262144, "available_providers": [ @@ -4308,6 +4309,7 @@ "flavor": "chat", "input_cost_per_mil_tokens": 2, "output_cost_per_mil_tokens": 6, + "deprecation_date": "2026-05-31", "parent": "mistral-large-latest", "fallback_models": [ "publishers/mistralai/models/mistral-large-2411" @@ -4347,6 +4349,7 @@ "multimodal": true, "input_cost_per_mil_tokens": 2, "output_cost_per_mil_tokens": 6, + "deprecation_date": "2026-05-31", "parent": "pixtral-large-latest", "max_input_tokens": 128000, "max_output_tokens": 128000, @@ -4361,6 +4364,8 @@ "input_cost_per_mil_tokens": 1.5, "output_cost_per_mil_tokens": 7.5, "displayName": "Mistral Medium 3.5", + "reasoning": true, + "reasoning_budget": true, "max_input_tokens": 256000, "available_providers": [ "mistral" @@ -4386,6 +4391,8 @@ "input_cost_per_mil_tokens": 0.15, "output_cost_per_mil_tokens": 0.6, "displayName": "Mistral Small", + "reasoning": true, + "reasoning_budget": true, "max_input_tokens": 262144, "max_output_tokens": 131072, "available_providers": [ @@ -4666,6 +4673,7 @@ "input_cost_per_mil_tokens": 0.59, "output_cost_per_mil_tokens": 0.79, "displayName": "Llama 3.3 70B Versatile 128k", + "deprecation_date": "2026-08-16", "max_input_tokens": 128000, "max_output_tokens": 32768, "available_providers": [ @@ -4724,6 +4732,7 @@ "output_cost_per_mil_tokens": 0.34, "displayName": "Llama 4 Scout (17Bx16E)", "experimental": true, + "deprecation_date": "2026-07-17", "max_input_tokens": 131072, "max_output_tokens": 8192, "available_providers": [ @@ -4776,6 +4785,8 @@ "input_cost_per_mil_tokens": 1.5, "output_cost_per_mil_tokens": 7.5, "displayName": "Mistral Medium 3.5", + "reasoning": true, + "reasoning_budget": true, "max_input_tokens": 256000, "available_providers": [ "mistral" @@ -6112,6 +6123,7 @@ "input_cache_read_cost_per_mil_tokens": 0.15, "displayName": "Gemini 3.5 Flash", "reasoning": true, + "reasoning_budget": true, "fallback_models": [ "publishers/google/models/gemini-3.5-flash" ], @@ -6778,7 +6790,7 @@ "reasoning_budget": true, "parent": "anthropic.claude-sonnet-4-6", "max_input_tokens": 1000000, - "max_output_tokens": 64000, + "max_output_tokens": 128000, "available_providers": [ "bedrock" ] @@ -6796,7 +6808,7 @@ "reasoning_budget": true, "parent": "anthropic.claude-sonnet-4-6", "max_input_tokens": 1000000, - "max_output_tokens": 64000, + "max_output_tokens": 128000, "available_providers": [ "bedrock" ] @@ -6813,7 +6825,7 @@ "reasoning": true, "reasoning_budget": true, "max_input_tokens": 1000000, - "max_output_tokens": 64000, + "max_output_tokens": 128000, "available_providers": [ "bedrock" ] @@ -8196,7 +8208,7 @@ "reasoning": true, "reasoning_budget": true, "max_input_tokens": 1000000, - "max_output_tokens": 64000, + "max_output_tokens": 128000, "available_providers": [ "vertex" ] @@ -9393,7 +9405,7 @@ "flavor": "chat", "input_cost_per_mil_tokens": 1.4, "output_cost_per_mil_tokens": 4.4, - "input_cache_read_cost_per_mil_tokens": 0.26, + "input_cache_read_cost_per_mil_tokens": 0.14, "displayName": "GLM 5.2", "reasoning": true, "max_input_tokens": 1048576, @@ -12297,5 +12309,19 @@ "available_providers": [ "vertex" ] + }, + "gemma-4-31b": { + "format": "openai", + "flavor": "chat", + "multimodal": true, + "input_cost_per_mil_tokens": 0.99, + "output_cost_per_mil_tokens": 1.49, + "displayName": "Gemma 4 31B", + "reasoning": true, + "max_input_tokens": 131072, + "max_output_tokens": 40960, + "available_providers": [ + "cerebras" + ] } }