Model (LLM)

A Model is an entry of the LLM catalog an Agent can point at through llm_model. It is not stored per workspace: GET /models (scope inbox/agents) returns the AI Gateway catalog narrowed to models that have a known price, so every model you can pick is costed and counted against the workspace monthly limit. Models without a price are never offered.

Schema

Field	Type	Description
id	string	Gateway id, provider/model. The value accepted in Agent.llm_model.
provider	string	First segment of id.
model	string	Second segment of id.
name	string	Display name.
description	string | null	Provider description, when available.
pricing.input / pricing.output	number	USD per token.
pricing.cache_read / pricing.cache_write	number	USD per token for prompt-cache reads and writes.
pricing.valid_from	datetime	When this price became effective.

pricing is null only if a price row disappears between the catalog fetch and the price lookup; in practice every listed model carries a price.

Example

{
  "data": [
    {
      "id": "anthropic/claude-sonnet-5",
      "provider": "anthropic",
      "model": "claude-sonnet-5",
      "name": "Claude Sonnet 5",
      "description": null,
      "pricing": {
        "input": 0.000003,
        "output": 0.000015,
        "cache_read": 0.0000003,
        "cache_write": 0.00000375,
        "valid_from": "2026-09-01T00:00:00.000Z"
      }
    }
  ]
}

Errors

502 gateway_error when the AI Gateway catalog cannot be fetched. Prices come from the local model_prices table and are synced by the operator; a model missing from the list usually means its price has not been synced yet.