Quick answer
OpenAI API rate limits Varies by model and usage tier. Check the organization limits page and response headers for the exact active RPM and TPM values rather than assuming one global number.
Verified Aug 22, 2026Official source
Current limits
| Constraint | Current value | Scope | Verified source |
|---|---|---|---|
| API rate-limit scopeOpenAI measures RPM, RPD, TPM, TPD, IPM, and certain audio limits; the first exhausted limiter applies. | Varies by model and usage tier | All API organizations | OpenAIAug 22, 2026 |
Why does this limit matter?
A request can exhaust a request-count limiter while token capacity remains, or exhaust token capacity before the request count is reached.
Limits are shared at organization and project scope, so one worker or teammate can consume capacity observed by another client.
What should you check?
- Open the Limits section in the OpenAI developer console for the organization and project.
- Inspect x-ratelimit-limit-requests, x-ratelimit-limit-tokens, remaining, and reset headers on API responses.
- Treat Retry-After as the minimum wait for eligible temporary 429 responses.
Important caveats
- Model families can share rate-limit pools, and long-context requests can have separate constraints.
- The numeric examples in response-header documentation are examples, not universal allowances for every account.
HyperObserve reports the documented platform constraint. Your application, SDK, gateway, provider, region, or account can impose a lower effective limit.
Related