Skip to main content

Currency and units

Provider tables show input, output, cache read, and cache write rates in USD per million tokens. The Pricing API returns these token rates in USD per token. Multiply a rate by 1,000,000 to match the tables. Other price fields may use different billing units. The catalog does not assign guessed units to them.

Missing prices

A dash (—) represents a source rate of zero or an unavailable column. The source does not reliably distinguish an unavailable rate from a free rate. Do not assume a dash means free inference. The main table includes models with known public rates. Enterprise-only models appear in a collapsed table. Other models without prices are omitted. When the API returns both a priced record and a null duplicate for the same provider and model, the catalog keeps the priced record.

Public and Enterprise pricing

The sidebar separates providers with known public rates from providers whose model prices are not public. Public provider pages include cURL request and response examples. Providers without known rates show metadata and an Enterprise contact banner. Contact us to get an API key for Enterprise pricing.

Exact identifiers

Provider pages show priced models and models explicitly marked as enterprise-only. Prices join metadata using the exact provider_id and model_id. Provider prefixes, case differences, and aliases remain intact. Missing metadata does not hide a price. Select a model ID in a provider’s price table to retrieve that provider/model pair from the API. To request the same exact model ID across providers, query /v1/prices with model_id and omit provider_id. The API does not merge differently named model aliases.

Refresh frequency

The catalog targets an hourly refresh at 17 minutes past the hour. API availability and validation can delay updates. Failed or incomplete runs retain the published site. Mintlify displays modification dates from each page’s Git history. The three API endpoints do not expose a shared snapshot version. Retrieval is not transactionally consistent. Use the Pricing API quickstart to retrieve current prices and calculate request costs.
Last modified on October 29, 2026