Get credits

Estimates what a request would cost without sending it or charging the account. Send the body you would send to the model's endpoint. Video models are priced exactly as `POST /v1/videos` freezes the job price, from `seconds`, `size` and `media`; text to speech is priced by the characters of `input`; chat models get an upper bound from the text of `messages` and `max_tokens`. The model string is resolved the same way a request is, at the caller's effective price (including contract pricing); when several routes could serve it at different prices, the most expensive is returned, because the serving route depends on availability at request time. Time-window pricing (peak/off-peak) is resolved at the current UTC time.