Models & providers

Coco Router registers 1,300+ model entries. They come in three kinds:

1. Named models

Concrete entries used by the router and for non-chat endpoints: gpt-4o-mini, gpt-4o, gpt-4-turbo, o1, whisper-1, gpt-4o-transcribe, text-embedding-3-small, text-embedding-3-large, text-embedding-ada-002, and the virtual router models auto, smart-ai, default.

2. Provider wildcards

Every provider below is wired as a provider/* wildcard. All of that provider's models are callable as provider/model-name the moment its API key is present in the server .env — including models the provider releases in the future.

Callable asEnv keyStatus
openai/*OPENAI_API_KEYactive
gemini/*GEMINI_API_KEYactive
openrouter/* (343 models)OPENROUTER_API_KEYactive
anthropic/*ANTHROPIC_API_KEYkey pending
groq/*GROQ_API_KEYkey pending
mistral/*MISTRAL_API_KEYkey pending
deepseek/*DEEPSEEK_API_KEYkey pending
xai/*XAI_API_KEYkey pending
cohere/*COHERE_API_KEYkey pending
perplexity/*PERPLEXITYAI_API_KEYkey pending (Perplexity available via OpenRouter)
together_ai/*TOGETHERAI_API_KEYkey pending
fireworks_ai/*FIREWORKS_AI_API_KEYkey pending
deepinfra/*DEEPINFRA_API_KEYkey pending
cerebras/*CEREBRAS_API_KEYkey pending
sambanova/*SAMBANOVA_API_KEYkey pending
ai21/*AI21_API_KEYkey pending
nvidia_nim/*NVIDIA_NIM_API_KEYkey pending
voyage/*VOYAGE_API_KEYkey pending
bedrock/*AWS credentialswired — see Amazon Bedrock

3. The OpenRouter catalog

The single OpenRouter key unlocks its full catalog (343 models at last count) — Meta Llama, Google Gemini, Anthropic Claude, DeepSeek, Qwen, Mistral, Perplexity Sonar (live web search), free community models, and more — all as openrouter/author/model.

Activating a new provider

Provider keys live in the gateway's server configuration. Adding one is an administrator task — once the key is in place, the provider's entire catalog goes live and can be used directly or placed in an Auto-Switch tier pool. See Managing the stack.