Model calls
Per token
Charged at the catalogue rate for the model that actually served the request. Nothing is added on top on the way out.
- Input and output metered separately
- Rate shown before the call, not after
- Every request itemised in the usage log