Billing and usage

Charges depend on the selected model, the provider that served the request and actual usage. Check current model-detail prices and request-level billing records rather than treating documentation examples as fixed prices.

Billing units

A model may bill input/output tokens, cache usage, image or video units, audio units, or a fixed request charge. Only the meters that apply to the selected model are charged. A per-request price appears on the output side with - on the input side; it is not charged twice.

Token and media usage is measured in memory. Usage records retain billing metadata, not prompt or completion content. Personal requests use wallet balance; team requests are governed by the project and key budgets.

Review costs

Compare request volume, token usage, retries, model/provider, latency and request cost in the console. Separate development, production and Agent keys to identify unexpected usage.

Each user has one wallet with an immutable currency, fixed at registration: USD. Changing the interface language does not change the account currency. Wallets are topped up through Stripe only. Mapping rates are configured independently in both currencies, with no conversion or transfers between them. Anonymous prices follow the currency of this site; signed-in prices follow the account currency.

Price-first routing compares providers for the same model and never substitutes a different model. See routing and budgets.