From 543d9290dbf4f9c6c5e6f4a5878a6a119a9222f9 Mon Sep 17 00:00:00 2001 From: erin2722 <16248113+erin2722@users.noreply.github.com> Date: Sat, 25 Jul 2026 10:42:58 +0000 Subject: [PATCH] chore: update model catalog from bot issues --- packages/proxy/schema/index.ts | 9 ++ packages/proxy/schema/model_list.json | 147 ++++++++++++++++++++++++-- 2 files changed, 150 insertions(+), 6 deletions(-) diff --git a/packages/proxy/schema/index.ts b/packages/proxy/schema/index.ts index 91eeb57a..5da3048a 100644 --- a/packages/proxy/schema/index.ts +++ b/packages/proxy/schema/index.ts @@ -1114,6 +1114,15 @@ export const AvailableEndpointTypes: { [name: string]: ModelEndpointType[] } = { "command-r-plus-08-2024": ["cohere", "openrouter"], "command-r-08-2024": ["cohere", "openrouter"], "command-r7b-12-2024": ["cohere", "openrouter"], + "databricks-gemini-3-6-flash": ["databricks"], + "databricks-glm-5-2": ["databricks"], + "databricks-inkling": ["databricks"], + "google.gemma-3-12b-it": ["bedrock"], + "google.gemma-3-4b-it": ["bedrock"], + "nvidia.nemotron-nano-9b-v2": ["bedrock"], + "openai.gpt-5.6-sol": ["bedrock"], + "openai.gpt-5.6-terra": ["bedrock"], + "openai.gpt-5.6-luna": ["bedrock"], }; const modelEndpointAvailableModels = getAvailableModels(); diff --git a/packages/proxy/schema/model_list.json b/packages/proxy/schema/model_list.json index 8f7ea3ba..12b25bf2 100644 --- a/packages/proxy/schema/model_list.json +++ b/packages/proxy/schema/model_list.json @@ -27,6 +27,9 @@ "input_cache_write_cost_per_mil_tokens": 1.25, "displayName": "GPT-5.6 Luna", "reasoning": true, + "fallback_models": [ + "openai.gpt-5.6-luna" + ], "max_input_tokens": 1050000, "max_output_tokens": 128000, "available_providers": [ @@ -45,6 +48,9 @@ "input_cache_write_cost_per_mil_tokens": 6.25, "displayName": "GPT-5.6 Sol", "reasoning": true, + "fallback_models": [ + "openai.gpt-5.6-sol" + ], "max_input_tokens": 1050000, "max_output_tokens": 128000, "available_providers": [ @@ -63,6 +69,9 @@ "input_cache_write_cost_per_mil_tokens": 3.125, "displayName": "GPT-5.6 Terra", "reasoning": true, + "fallback_models": [ + "openai.gpt-5.6-terra" + ], "max_input_tokens": 1050000, "max_output_tokens": 128000, "available_providers": [ @@ -7898,7 +7907,11 @@ "flavor": "chat", "input_cost_per_mil_tokens": 0.1, "output_cost_per_mil_tokens": 0.3, - "parent": "mistral-small-latest" + "deprecation_date": "2025-11-30", + "parent": "mistral-small-latest", + "available_providers": [ + "mistral" + ] }, "ministral-8b-2410": { "format": "openai", @@ -7919,14 +7932,22 @@ "flavor": "chat", "input_cost_per_mil_tokens": 0.2, "output_cost_per_mil_tokens": 0.6, - "displayName": "Mistral Saba" + "displayName": "Mistral Saba", + "deprecation_date": "2025-09-30", + "available_providers": [ + "mistral" + ] }, "mistral-saba-2502": { "format": "openai", "flavor": "chat", "input_cost_per_mil_tokens": 0.2, "output_cost_per_mil_tokens": 0.6, - "parent": "mistral-saba-latest" + "deprecation_date": "2025-09-30", + "parent": "mistral-saba-latest", + "available_providers": [ + "mistral" + ] }, "llama3-3-70b": { "format": "openai", @@ -8128,11 +8149,10 @@ "reasoning_budget": true, "fallback_models": [ "anthropic.claude-opus-5", + "eu.anthropic.claude-opus-5", "global.anthropic.claude-opus-5", "publishers/anthropic/models/claude-opus-5", - "us.anthropic.claude-opus-5", - "eu.anthropic.claude-opus-5", - "au.anthropic.claude-opus-5" + "us.anthropic.claude-opus-5" ], "max_input_tokens": 1000000, "max_output_tokens": 128000, @@ -8858,6 +8878,32 @@ "bedrock" ] }, + "google.gemma-3-12b-it": { + "format": "openai", + "flavor": "chat", + "multimodal": true, + "input_cost_per_mil_tokens": 0.09, + "output_cost_per_mil_tokens": 0.29, + "displayName": "Gemma 3 12B IT", + "max_input_tokens": 128000, + "max_output_tokens": 8000, + "available_providers": [ + "bedrock" + ] + }, + "google.gemma-3-4b-it": { + "format": "openai", + "flavor": "chat", + "multimodal": true, + "input_cost_per_mil_tokens": 0.04, + "output_cost_per_mil_tokens": 0.08, + "displayName": "Gemma 3 4B IT", + "max_input_tokens": 128000, + "max_output_tokens": 8000, + "available_providers": [ + "bedrock" + ] + }, "zai.glm-5": { "format": "converse", "flavor": "chat", @@ -8969,6 +9015,48 @@ "bedrock" ] }, + "openai.gpt-5.6-sol": { + "format": "openai", + "flavor": "chat", + "multimodal": true, + "input_cost_per_mil_tokens": 5.5, + "output_cost_per_mil_tokens": 33, + "input_cache_read_cost_per_mil_tokens": 0.55, + "displayName": "OpenAI GPT-5.6 Sol (Bedrock)", + "reasoning": true, + "max_input_tokens": 272000, + "available_providers": [ + "bedrock" + ] + }, + "openai.gpt-5.6-terra": { + "format": "openai", + "flavor": "chat", + "multimodal": true, + "input_cost_per_mil_tokens": 2.75, + "output_cost_per_mil_tokens": 16.5, + "input_cache_read_cost_per_mil_tokens": 0.28, + "displayName": "OpenAI GPT-5.6 Terra (Bedrock)", + "reasoning": true, + "max_input_tokens": 272000, + "available_providers": [ + "bedrock" + ] + }, + "openai.gpt-5.6-luna": { + "format": "openai", + "flavor": "chat", + "multimodal": true, + "input_cost_per_mil_tokens": 1.1, + "output_cost_per_mil_tokens": 6.6, + "input_cache_read_cost_per_mil_tokens": 0.11, + "displayName": "OpenAI GPT-5.6 Luna (Bedrock)", + "reasoning": true, + "max_input_tokens": 272000, + "available_providers": [ + "bedrock" + ] + }, "xai.grok-4.3": { "format": "openai", "flavor": "chat", @@ -9273,6 +9361,18 @@ "bedrock" ] }, + "nvidia.nemotron-nano-9b-v2": { + "format": "converse", + "flavor": "chat", + "input_cost_per_mil_tokens": 0.06, + "output_cost_per_mil_tokens": 0.23, + "displayName": "NVIDIA Nemotron Nano 9B v2", + "max_input_tokens": 128000, + "max_output_tokens": 8000, + "available_providers": [ + "bedrock" + ] + }, "amazon.nova-2-lite-v1:0": { "format": "converse", "flavor": "chat", @@ -12722,6 +12822,41 @@ "databricks" ] }, + "databricks-gemini-3-6-flash": { + "format": "openai", + "flavor": "chat", + "multimodal": true, + "displayName": "Gemini 3.6 Flash", + "reasoning": true, + "reasoning_budget": true, + "max_input_tokens": 1048576, + "max_output_tokens": 65536, + "available_providers": [ + "databricks" + ] + }, + "databricks-glm-5-2": { + "format": "openai", + "flavor": "chat", + "displayName": "GLM 5.2", + "reasoning": true, + "max_input_tokens": 262144, + "max_output_tokens": 128000, + "available_providers": [ + "databricks" + ] + }, + "databricks-inkling": { + "format": "openai", + "flavor": "chat", + "multimodal": true, + "displayName": "Inkling", + "reasoning": true, + "max_input_tokens": 1048576, + "available_providers": [ + "databricks" + ] + }, "thinkingmachines/inkling": { "format": "openai", "flavor": "chat",