Pricing and billing
Understand public meter prices, prepaid reservations, measured settlement, and content-blind usage history.
Public prices are meter-specific
Model detail pages show the customer price for each published meter, normally expressed per one million units. Input, cached input, output, and other meters can differ. The public price is what determines your customer charge; provider route selection and internal cost are private operational data.
Requests reserve before they run
The final meter count is unknown at the start of an inference request. To keep a prepaid balance safe under concurrency, OurToken calculates a bounded estimate and places that amount on hold. The held amount reduces available credit but is not yet a final charge.
After the response returns, OurToken settles the measured usage against the effective price snapshot for that request and releases the unused portion of the reservation. Failed validation or delivery releases the hold rather than charging an estimate.
Reconcile from the ledger
Dashboard usage and exports expose the model, normalized meters, price snapshot, final charge, time, API-key reference, and lifecycle status. Prompts, completions, provider credentials, and private routing details do not belong in that customer billing record.
Why a snapshot matters
Changing a catalog price later does not rewrite an old request. Each settled charge remains explainable using the price that applied when the request ran.