Docs / Reference

Models

The public models endpoint is GET /api/inference/models. Use ?type=llm, ?type=media, or ?type=both. If omitted, it defaults to llm.

LLM models

GET /api/inference/models?type=llm returns pricing and input types. Costs are per million tokens.

{
  "type": "llm",
  "models": [
    {
      "id": "gemma-4-31b",
      "name": "Gemma 4",
      "input_cost_per_million": 0.13,
      "output_cost_per_million": 0.38,
      "input_types": ["text", "image"]
    }
  ]
}

Media models

GET /api/inference/models?type=media returns only id and name. Costs depend on several factors and are obtained by filling in the desired parameters on the respective media generation pages.

{
  "type": "media",
  "models": [
    {
      "id": "gpt-image-1.5",
      "name": "GPT Image 1.5"
    }
  ]
}