API DocsAPI Docs
Rate limits and concurrency
Per-key request limits and account job limits
Limits protect service availability and are applied before expensive work.
| Limit | Current value |
|---|---|
| Job creation | 20 requests per minute per API key, across all create endpoints |
| Authenticated reads | 120 requests per minute per API key |
| Active jobs | 20 queued or processing jobs per account |
| Batch count | Up to 4 outputs per generation or edit job |
| Bounded wait | Up to 30 seconds |
Exceeding a request or active-job limit returns HTTP 429 with
RATE_LIMIT_EXCEEDED. Use exponential backoff with jitter and respect
Retry-After when it is present. Idempotency keys must remain unchanged while
retrying the same logical request.
The active-job limit is account-wide, even when multiple API keys are used. Completed jobs do not count toward it.
At the maximum batch size, 20 active generation or edit jobs can request up to 80 outputs in total. This is a submission capacity limit, not a guarantee that every output runs simultaneously.