dotsfeed
← Updates

Reset

Codex weekly allowance crashed on Oct 16? Note your numbers and file a ticket before you spend a banked reset

Observed· Oct 17, 2026Published Oct 17, 2026

Dozens of Codex users report their weekly allowance jumping toward 100% used in minutes around 11:10 AM PT on Oct 16, with almost no usage behind it. OpenAI hasn't explained it publicly yet. Here's how to check whether you were hit, what to record, and where to report it.

What changed

On Oct 16, many Codex users saw their weekly allowance jump almost to empty in seconds or minutes, with little or no work running. The clearest logs put the jumps between about 11:08 and 11:12 AM PT (18:08 to 18:12 UTC). One person went from 77% left to 10% left in 38 seconds while their local logs recorded only about 381,000 tokens, nearly all of it cached input. Others report 24% to 97% used in 12 seconds, or a laptop that was closed overnight losing a big chunk of its allowance.

The main GitHub issue (#54677) now has more than 30 reports from macOS, Windows, and Linux, across the desktop app and the CLI. One reporter shared an OpenAI support reply saying server-side records show "the weekly allowance used to calculate your meter decreased while the amount already consumed initially stayed the same." In other words, the size of their weekly allowance shrank, not their usage. That reply came from OpenAI's AI support agent, not from the engineering team.

Who is affected

Mostly ChatGPT Pro users, and the reports that include plan details are mostly Pro 5x (shown as prolite in Codex logs). One commenter noticed that the affected accounts they compared all had a weekly reset at nearly the same moment, around 11:13 PM PT on Oct 21 (06:13 UTC Oct 22), which points to a server-side event hitting one group of accounts rather than anything users did. Some affected users had recently downgraded from Pro $200 to Pro $100, but many say they hadn't changed plans at all, so a plan change doesn't explain every case.

What to do now

  • Check Settings > Usage & billing in Codex, or /status in the CLI, and compare it with what you remember from Oct 15 or 16.
  • If your number looks wrong, write down your plan, the before and after percentages, and roughly when it changed. If you're comfortable digging, the rate_limits entries in your local Codex session logs show used_percent, plan_type, and resets_at, which is what other reporters are comparing.
  • Open a ticket at help.openai.com and mention GitHub issue #54677. You can add your timeline there too, but don't post account emails or other private details.
  • Think before you spend a banked reset. Several people say their allowance drained much faster than usual right after using one, so if you can wait for an answer, waiting may be safer. If you're blocked on real work, it's your call.
  • If you have extra usage credits, those still work, but using them doesn't restore what you lost.

What is not confirmed

OpenAI hasn't publicly explained the drop, said whether it was a mistake, or promised to restore anything, and there is no incident on the OpenAI status page. The timelines come from users' own logs and Codex-assisted log summaries posted on GitHub and X, and they haven't been independently verified. The faster-drain-after-reset reports are also field reports. DotsFeed hasn't reproduced any of this, and we'll update this piece if OpenAI responds.

Sources

Sources

Get updates like this every morning

  1. ① Email
  2. ② Card on Stripe
  3. ③ 7 days free

Then $2/month · cancel anytime in one click