Models

Model identifiers and capabilities.

lmchat uses a provider/model naming convention and exposes a models endpoint for discovery. Routing policies can map a logical model to specific providers.

Model naming

Convention

Use identifiers like anthropic/claude, openai/gpt, google/gemini.

{
  "model": "anthropic/claude",
  "messages": [{"role":"user","content":"Hello"}]
}

List models

GET /api/v1/models

Response is an OpenAI-style list envelope where possible.

{
  "object": "list",
  "data": [
    {
      "id": "anthropic/claude",
      "object": "model",
      "created": 1710000000,
      "owned_by": "anthropic",
      "capabilities": {
        "tools": true,
        "vision": false,
        "json_mode": true
      }
    }
  ]
}

Choosing a model

Start with a general-purpose reasoning model for mixed workloads.
Use smaller/faster models for high-QPS, short outputs, or classification.
Use tool-capable models for agents and workflows.
If you care about uptime, configure routing & fallbacks.