Multiple paid Codex accounts affected: limits drain after one prompt, company monthly credits burned in one day
This is no longer isolated to one account. It now looks like a systemic, billing-impacting rate-limit / usage-accounting regression across multiple paid Codex accounts.
Related reports I opened earlier:
Previous Feedback ID:
019eaf5c-26e2-79a0-b678-ffa25afc0863
What I am seeing now:
- My personal paid Codex account is still affected. Limits drain far faster than actual usage.
- A separate work Codex account is now showing the same behavior: limits can disappear after a single question.
- On the work/company account, even company credits were affected: a monthly company budget was consumed in one day.
- Another Plus account in my household started showing a similar fast limit-drain pattern today after lunch.
This pattern started for me around mid June, briefly improved, then regressed again. It is now multiplying across accounts, not staying isolated to one user, one plan, or one local session.
This needs urgent attention because it is now a paid-product and billing-impacting issue:
- Paid limits are being consumed much faster than the work performed.
- Credits/monthly budgets can be exhausted in a day.
- Users cannot reliably use Codex even while paying for it.
- There are already many similar reports in issues, but the problem still appears unresolved.
I understand there may be anti-abuse, routing, or rate-limiter systems involved, but from the user side this looks like false-positive or broken usage attribution. I also completed identity/passport verification, and the issue still persists.
Please escalate this to the team responsible for quota accounting / rate limiting / billing impact. At minimum, please:
- publicly acknowledge whether this is a known ongoing incident or regression
- investigate why multiple independent paid accounts are seeing the same abnormal drain
- check for account-level, organization-level, IP/routing, or anti-abuse false-positive flags
- restore incorrectly drained limits and company credits for affected accounts
- stop the drain while the investigation is ongoing
Environment for this report:
- Codex CLI version:
codex-cli 0.142.5 - App: Codex Desktop
- OS: macOS
15.7.4(24G517) - Current region / connection: Poland, P4 Sp. z o.o. (
AS9141) - Date observed again: July 9, 2026 CEST
I can provide private account details, company account details, screenshots, and fresh feedback IDs through a private support channel. I am not posting account emails or identity-verification details in a public GitHub issue.
6 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
Also hitting this. Demanded I use a /usage reset. Did that, issued two failed commands due to shell tool unavailability then asked for another reset.
Update from 2026-07-09 02:17 CEST:
The state just briefly looked fixed: the limits reset down to around 0-1%. But about 5 minutes later they reverted back to the broken/drained state.
It also looks worse now than before. On the $100/month paid subscription, the entire 5-hour window can now disappear in about 20 minutes. This is not consistent with actual usage and looks like unstable quota/accounting state rather than normal model consumption.
Please preserve/investigate telemetry around this reset-and-rollback window. This strongly suggests the issue is still active and recurring, not resolved.
Still the same for me, too. I am using Pro x5 and the peroid starting from June 26th is my LOWEST since my account has been created and my weekly usage is gone within a single day. You can clearly see it in the graph below.
Unfortunately this issue is being ignored completely and overshadowed by the GPT5.6 release.
<img width="1249" height="429" alt="Image" src="https://github.com/user-attachments/assets/2886a72e-ecf0-467c-a855-5de9f2143864" />
Same here - noticed last night. The issue still persists this morning.
I can reproduce the same abnormal usage drain on a separate paid account.
Environment:
97% remaining in the secondary window
were running
The requests only discussed suspected abnormal Codex usage accounting.
They performed no repository-wide implementation or long-running task.
Yesterday, with GPT-5.5, including Very High reasoning effort,
more complex requests consumed many times less quota.