Reference

Rate limits

Per-workspace and per-model.

Default limits during beta:

Free tier: 60 req/min, 10 concurrent jobs
Basic: 120 / 20
Pro: 300 / 40
Creator: 600 / 80
Enterprise: custom

Exceeding returns 429 with Retry-After and X-RateLimit-Remainingheaders. Per-model concurrency is also enforced upstream at the provider; we retry-with-backoff on 429 from providers so your call doesn't see it unless it persists.

Last updated 2026-04-17Needs more detail? →