Billing basics
Billed by token usage; balance funds all keys while each key has its own spending limit.
The platform bills by token usage: each call deducts its token consumption at the model's unit price from the account balance. This page explains the billing logic and how to audit it.
How billing works
- The unit is tokens: the cost of a call is its input + output tokens, reported exactly in the response's
usagefield (see Chat completions); - Unit prices are per model: models differ in price — the billing type and price on each marketplace card are authoritative;
- Deducted from balance: all models draw from the same account balance, deducted in real time after each call.
Balance vs. key spending limits
They are independent and govern different things:
| Dimension | Scope | Where to set it |
|---|---|---|
| Account balance | The funding source for all models and all keys | Top up on the "Credits" page (see Top up and gift card redemption) |
| Key spending limit | Cumulative cap for a single key | Set when creating/editing a key on the "API Keys" page |
- Balance exhausted: all calls fail with an insufficient-balance error until you top up;
- Key limit reached: that key is refused, while other keys and the balance are unaffected. Topping up does not raise a key's limit.
Auditing consumption
- Account overview: balance and usage trend (by tokens, over time);
- Usage / logs pages: per-call model, token count, and cost, aggregatable over a time range;
- Reconciliation: compare your application's request counts against the per-request records on the logs page to locate discrepancies by status and timestamp.
If pricing changes, the marketplace cards and platform announcements are authoritative.