generated: '2026-09-18'
method: searched
source: https://thirds.ai/docs/rate-limits
limits:
- scope: per-api-key
  dimension: submission rate
  limit: 5 requests/second
  burst: idle burst of 21
- scope: per-api-key
  dimension: live jobs (queued or running)
  limit: 8
- scope: per-account
  dimension: live jobs (queued or running)
  limit: 16
  note: PDF and image work share account capacity
- scope: service
  dimension: paid queued-job backlog
  limit: 500
- scope: per-browser
  dimension: anonymous playground renders
  limit: 200 per month
response_headers:
- Retry-After
no_headers:
- X-RateLimit-*
exhaustion:
- status: 429
  code: rate_limited
  retry: 1-60s via Retry-After
- status: 429
  code: account_concurrency_limited
  retry: ~5s
- status: 429
  code: key_concurrency_limited
  retry: ~5s
- status: 429
  code: abuse_limited
  retry: ~3600s
- status: 503
  code: overloaded
  retry: ~30s
note: No remaining-request counter or X-RateLimit-* headers. Clients read Retry-After (seconds) and error.retry.retry_after_seconds.
  Accepted async jobs return 202 with a Location header (poll after ~1s).