Naming
Model IDs are always provider/model. Append :free for evaluation tiers. The ID never changes silently underneath you — a new revision gets a new ID.
Catalogue
Models & token prices
One endpoint, one key, one balance. Every price below is 5% below the market reference rate, in USD per one million tokens. Rates are synchronised with upstream pricing, and your dashboard always shows the exact rate applied to each request.
| Model | Provider | Context | Input / 1M | Output / 1M | Capabilities |
|---|
How these prices are set. We benchmark each model against the prevailing market
reference rate and sell 5% under it. When an upstream provider moves its price, ours follows within
one business day — prospectively, never retroactively against credit you have already bought.
Fallbacks
Pass a fallbacks array of model IDs. We try them in order when the primary is unavailable, and bill you only for the request that actually completed.
Deprecations
Upstream retirements are announced by email at least 30 days ahead where the provider gives us notice, with a recommended replacement ID.
Need a model that isn't listed?
We add upstreams on request, including self-hosted and private endpoints for enterprise customers.