Codex reports usage limit despite Pro plan and dashboard showing remaining 5h/weekly quota

Open 💬 1 comment Opened Jul 2, 2026 by Qicaoji7

What issue are you seeing?

I recently upgraded to ChatGPT Pro and the Pro plan appears active in the ChatGPT interface. However, Codex is blocking requests with a usage-limit message even though the Codex/usage panel still shows quota remaining.

Observed in the UI:

  • Plan: ChatGPT Pro
  • 5-hour usage window: 79% remaining
  • 5-hour reset time shown: 19:50
  • Weekly usage: 97% remaining
  • Weekly reset date shown: July 9
  • Model selected: GPT-5.5 / high-reasoning tier shown in the UI
  • Regular ChatGPT conversations still work normally

But when I try to use Codex, it reports that I have reached the usage limit and says I need to wait until the reset time.

This looks inconsistent because the dashboard/panel still shows available capacity, while Codex rate-limit enforcement blocks the request.

What steps can reproduce the bug?

  1. Use an account with an active ChatGPT Pro subscription.
  2. Open Codex from the ChatGPT web interface.
  3. Confirm that the usage panel shows remaining 5-hour and weekly quota.
  4. Try to send a Codex request.
  5. Codex blocks the request with a usage-limit message even though quota is still shown as available.

What is the expected behavior?

If the usage dashboard/panel shows remaining quota for the active Pro plan, Codex should allow the request, or the UI should clearly explain which separate quota bucket/model allowance/workspace limit is exhausted.

At minimum, the usage panel and the Codex rate-limit enforcement should be consistent.

Additional information

This may be related to recent quota/entitlement/rate-limit mismatch reports, for example:

  • #30041 — dashboard and /status showed remaining quota, but Codex still returned usage-limit errors.
  • #29243 — Pro 5x entitlement appeared to be treated as a lower-tier plan by the Codex rate limiter.
  • #30772 — after upgrade, Codex still reported usage limit.

I have also escalated this through OpenAI Support and can provide screenshots privately if needed. I am intentionally not attaching account email or unredacted screenshots in this public issue.

View original on GitHub ↗

This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗