API reference: Models

GET /v1/models lists the tiers your API key may use. Read it when your application starts, and choose tiers by what they can do, not by a fixed list.

GET https://dummydomain/v1/models
Authorization: Bearer YOUR_API_KEY

Response

{
  "object": "list",
  "data": [
    {
      "id": "Fast",
      "object": "model",
      "context_window": 49152,
      "max_output_tokens": 16384,
      "max_images_per_request": 8,
      "capabilities": {
        "chat": true,
        "streaming": true,
        "tools": true,
        "parallel_tool_calls": true,
        "vision": true,
        "response_format": ["text", "json_object", "json_schema"],
        "structured_outputs": true
      },
      "tool_context_window": 24576,
      "max_output_tokens_with_tools": 4096,
      "tool_choice_modes": ["none", "auto", "required", "named"],
      "tool_arguments_validated": true
    }
  ]
}

This example is shortened to one tier. Your server lists each tier it offers.

Fields

Field	Description
`id`	The tier name. Send it as `model`
`context_window`	Tokens the tier can handle in one request, prompt and answer together
`max_output_tokens`	The longest answer, in tokens
`max_images_per_request`	Images allowed in one request, on tiers that read images
`capabilities.tools`	`true` when the tier takes function tools
`capabilities.vision`	`true` when the tier reads images
`capabilities.response_format`	The `response_format` and `text.format` types it accepts
`capabilities.structured_outputs`	`true` when JSON schema output is checked against your schema
`tool_context_window`	The context window for requests that carry tools
`max_output_tokens_with_tools`	The longest answer when tools are in the request
`tool_choice_modes`	`none`, `auto`, `required`, and `named` for a forced function
`tool_arguments_validated`	`true`: function arguments are checked against your schema before you get them

Tiers and what they are good for are explained in Choose a tier.

Common questions

How do I know which ZenithAI tier supports tools or images?

Call GET /v1/models. Each tier has a capabilities object; tools and vision are true for the tiers on your server that take function calls and images. Read it at start-up instead of hard-coding tier names.