Models

List available virtual models and optionally real models from OpenRouter.

GET/v1/models

List all available models.

Request

curl https://antbase.ai/v1/models \
  -H "Authorization: Bearer YOUR_API_KEY"

Response

{
  "object": "list",
  "data": [
    {
      "id": "ant:auto",
      "object": "model",
      "created": 1708123456,
      "owned_by": "antbase",
      "description": "Balanced routing - ANT picks the best model"
    },
    {
      "id": "ant:fast",
      "object": "model",
      "created": 1708123456,
      "owned_by": "antbase",
      "description": "Optimized for speed"
    },
    {
      "id": "ant:balanced",
      "object": "model",
      "created": 1708123456,
      "owned_by": "antbase",
      "description": "Balanced speed and quality"
    },
    {
      "id": "ant:best",
      "object": "model",
      "created": 1708123456,
      "owned_by": "antbase",
      "description": "Maximum quality"
    }
  ]
}

Virtual Models

Virtual models are routing presets. When you request a virtual model, ANT classifies your prompt and routes to the optimal real model.

Model ID	Speed	Quality	Cost Priority	Best For
`ant:auto`	50	50	50	General purpose
`ant:fast`	100	30	70	Real-time, simple queries
`ant:balanced`	50	60	50	Business applications
`ant:best`	20	100	20	Complex reasoning, coding

Real Models

You can also request specific models by name to bypass ANT's routing. Any model supported by OpenRouter is available.

Popular Models

Model ID	Provider	Notes
`anthropic/claude-opus-4-6`	Anthropic	Most capable Claude
`anthropic/claude-3-5-sonnet-20241022`	Anthropic	Good balance
`anthropic/claude-haiku-4-5`	Anthropic	Fast and cheap
`openai/gpt-4o`	OpenAI	Multimodal
`openai/gpt-4o-mini`	OpenAI	Budget option
`google/gemini-2.0-flash-001`	Google	Very fast

See openrouter.ai/models for the complete list of 2,500+ models.

How Model Selection Works

Virtual model requested: ANT classifies your prompt and scores all real models based on the virtual model's preference weights.
Real model requested: ANT forwards directly to that model (still tracks usage and applies rate limits).
Model not found: Returns 400 error with code INVALID_MODEL.