Documentation

Models

The public catalog contains two Klara Base previews. They are smaller meantime versions while the new, more powerful Klara models are being finished. This table is live: disabled or unhealthy routes remain visible but cannot accept requests.

Public model catalog

Model	Behavior	Tier	Context	Input / M	Output / M	Request minimum	Status
Klara Base Preview v1	Direct, fast research and technical work while the next generation of Klara models is being completed.	Standard	202.8k	$2.50	$8.00	Set at launch	Temporarily unavailable
Klara Base Preview v2	Harder reasoning, coding, and research work with capability retention prioritized.	Standard	202.8k	$2.50	$8.00	Set at launch	Temporarily unavailable

The API identifies each model by its wire id. Pass one of these as the model field:

model-ids

# Wire ids for the "model" field
Klara Base Preview v1 ->  Klara-base-preview-v1  (direct early preview)
Klara Base Preview v2 ->  Klara-base-preview-v2  (capability-preserving preview)

Klara Base Preview v1

A smaller, direct preview for early research and technical work while the next generation of Klara models is being completed.

Klara Base Preview v2

A smaller preview tuned to retain capability on harder reasoning, coding, and research work while the more powerful Klara models are still being finished.

Choosing between them

Choose Klara Base Preview v1 for the more direct early preview, or v2 when capability retention on harder reasoning, coding, or research work matters more. Check the table for each route's live availability, context, price, and request minimum before sending.

What the models support

Context limits are published per route in the live table above.
Streaming over SSE and tool calls. See Streaming.
Text only. These models do not accept images.
30-day operational retention. Prompts and completions are never used to train models.

Access

Each Klara Base Preview is a standard paid route: use a standard-scoped key, sufficient credits, and an available release. Research Mode is only for approved higher-risk project scopes; those calls require a named, active entitlement, but approval never overrides a disabled or unhealthy route. Apply on the Research Mode page when that additional scope is needed.

Passing a model id

Set the model field to one of the wire ids. To switch models, change only that field:

bash

curl https://api.advancedmind.ai/v1/chat/completions \
  -H "Authorization: Bearer $AMAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "Klara-base-preview-v1",
    "messages": [{ "role": "user", "content": "..." }]
  }'

The id you pass is echoed in the response and recorded against your usage. You can also fetch the live catalog from GET /v1/models, which returns the same ids in the OpenAI list format.