Models¶

Blaick exposes many models from many providers behind one consistent request shape. You choose a model by its id and pass it to chat completions — the request body is identical no matter which provider ultimately serves it.

List available models¶

GET https://api.blaick.ai/api/v1/models

curl https://api.blaick.ai/api/v1/models \
  -H "Authorization: Bearer $BLAICK_API_KEY"

The catalog is filtered to what your account can use. On a trial, you'll only see trial-eligible models.

{
  "models": [
    {
      "id": "3f1c2b9a-0000-4000-8000-abcdefabcdef",
      "name": "Claude Sonnet 4",
      "provider": "anthropic",
      "model_id": "claude-sonnet-4-20250514",
      "tier": 3,
      "cost_multiplier": 2.0,
      "max_context_length": 200000,
      "supports_streaming": true,
      "is_trial_eligible": false,
      "is_available": true,
      "description": "Balanced model for most tasks."
    }
  ]
}

Field	Description
`id`	The UUID you pass as `model_id` in a completion.
`name`	Human-friendly display name.
`provider`	Upstream provider — `anthropic`, `openai`, `groq`, `bedrock`, `vertex`, and others.
`model_id`	The provider's own model identifier. Also accepted as `model_id` in a completion.
`tier`	Capability/price tier — higher generally means more capable and more expensive.
`cost_multiplier`	Multiplier applied to token usage for billing. `cost_tokens = total_tokens × cost_multiplier`.
`max_context_length`	Maximum combined input + output tokens.
`supports_streaming`	Whether the model can stream.
`is_trial_eligible`	Whether trial accounts may use it.
`is_available`	Whether it's currently online.

Get one model¶

curl https://api.blaick.ai/api/v1/models/{model_id} \
  -H "Authorization: Bearer $BLAICK_API_KEY"

Choosing a model¶

Match capability to the task. Reserve high-tier models for hard reasoning; use lower tiers for classification, extraction, and short answers.
Mind the cost_multiplier. Two models can report the same raw token usage but bill differently. The multiplier is applied to arrive at cost_tokens. See Usage & Tokens.
Check max_context_length before sending long inputs.
Let Blaick decide. If you're unsure, set auto_route: true on the completion and Blaick picks the cheapest capable model for each prompt. See Auto-routing.

Referencing a model

Both the model's id (UUID) and its model_id (e.g. claude-sonnet-4-20250514) are accepted in the model_id field of a completion request. The UUID is the most stable reference.