Errors & limits

Errors follow the OpenAI error shape: an error object with message, type and code fields, plus a conventional HTTP status.

StatusMeaning
400Bad request — a parameter is malformed or missing. The body explains which.
401Missing or invalid API key.
402Insufficient balance — top up and retry.
404Unknown model slug or endpoint.
408Upstream timeout — safe to retry; you are not billed for failed generations.
429Rate limited — back off and retry. Limits scale with account history.
502 / 503Upstream provider error — the gateway retries automatically across providers; persistent failures are safe to retry.

Rate limits

Default limits are generous for development and rise automatically with sustained usage. If you are planning a burst — batch jobs, launches, evals — contact us and we will raise the ceiling ahead of time.