Models

Discover the public tier catalog at runtime with GET /api/v1/models.

GET /api/v1/models returns the public tier catalog (the tiers you can pass as model, with their capabilities), so you read it at runtime instead of hardcoding tier strings.

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

const res = await fetch('https://api.cruxy.ai/api/v1/models', {
  headers: { Authorization: `Bearer ${process.env.CRUXY_API_KEY}` },
})
 
const { data } = await res.json()
for (const model of data) {
  console.log(model.id, model.context_window, model.supports_tools)
}

Tiers, never provider models

The catalog lists Cruxy's public tiers (kavi, vaani, and mira) in the OpenAI discovery shape. Each entry's id is the tier string you pass as model to Chat. The response never exposes the upstream provider model behind a tier, so a tier is a stable id you can depend on even when the backing model changes. Choosing a tier compares the tiers and when to use each.

	Kavi	Vaani	Mira
Model id	`kavi`	`vaani`	`mira`
What it is	The most capable tier.	The balanced tier.	The fastest tier.
Best for	The hardest tasks: deep reasoning and the largest output budget.	Everyday calls: strong quality at lower latency and cost.	Short, high-volume calls where latency and cost matter most.

Response shape

The envelope is a list of model objects:

{
  "object": "list",
  "data": [
    { "object": "model", "id": "kavi", "label": "..." }
  ]
}

Each entry describes one tier:

Field	Meaning
`id`	The tier string to pass as `model`, e.g. `kavi`.
`label`	Human-readable name for the tier.
`object`	Always `"model"`.
`context_window`	Maximum input size, in tokens.
`max_output`	Maximum tokens the tier will generate.
`latency_target_ms`	The tier's latency target, in milliseconds.
`supports_tools`	Whether the tier supports tool calling; see Define tools.
`pricing`	`currency`, `input_unit`, and `output_unit`, the per-unit prices.

Read the concrete values from the live response rather than assuming them; they can change as tiers evolve.

Errors

GET /api/v1/models returns 401 if the key is missing or invalid. See Errors.

Next steps

Chat: use a tier id as model.
Choosing a tier: pick the tier for each call.
Define tools: for tiers where supports_tools is true.
Changelog: when the catalog's metadata changed.