Skip to content

Limits – Reduced usage after 20 minutes. #15073

Closed as duplicate
Closed as duplicate

Description

@YoloMark98

What version of Codex CLI is running?

codex-cli 0.115.0

What subscription do you have?

Pro

Which model were you using?

gpt-5.4 high

What platform is your computer?

Linux 6.8.0-106-generic x86_64 x86_64

What terminal emulator and version are you using (if applicable)?

Tabby

What issue are you seeing?

Today, my credits on my account user-grEQqq7dokHb56wxlYbBeUWO reset correctly at 5:50 PM CEST and all usage limits were back at 100%. However, while I was running a task, I noticed it was consuming far more than normal. Specifically, from around 6:10 PM CEST to about 6:30 PM CEST, so within roughly a 30-minute window, I used up all of the 5-hour limits and 30% of the weekly limits.

The x2 promo has not ended yet, so I do not think that is the reason. This concerns me even more for when it does end, because at this rate the Pro account will basically become the new Plus account.

I have attached the thread from my session, and I hope you can review it to identify the issue that occurred, because it is clear that there is some kind of problem. I have never consumed this much in 30 minutes, not even when pushing it to the maximum.

What steps can reproduce the bug?

Uploaded thread: 019d01f1-2bc3-75a0-a802-5b2c4c585cb6

What is the expected behavior?

No response

Additional information

Image

As you can see from the photo, the usage section doesn’t show the full peak in usage either, whereas the main section does:

Image

Activity

  1. added
    CLIIssues related to the Codex CLI
    rate-limitsIssues related to rate limits, quotas, and token usage reporting
    on Mar 18, 2026
  2. etraut-openai commented on Mar 18, 2026

    @etraut-openai
    Contributor

    Please refer to #14593.

  3. Sokalledcoder commented on Mar 18, 2026

    @Sokalledcoder

    I am experiencing the same issue described above, and I have seen it on both Plus and Business accounts, not just Pro.

    In my case, usage limits appear to be dropping unusually fast. I can send only a small number of prompts, sometimes as few as five, and watch the weekly limit decrease prompt by prompt. Because of how consistently this is happening, it seems like there may be a real issue after version v0.115.0 rather than isolated user confusion.

    Another unusual detail on my account is that my usage breakdown was categorized as “Other,” even though I only use the Codex CLI on Linux and Windows 10. In the past, the exact same workflow was always recorded as CLI, so this change is unexpected and may be related.

    I am mentioning this both to confirm that another user is seeing the same behavior and to encourage a closer investigation. Referring people back to threads from several days ago does not really address what appears to be an active and wider problem. At this point, it seems likely that multiple users are being affected.

  4. YoloMark98 commented on Mar 18, 2026

    @YoloMark98
    Author

    Please refer to #14593.

    If you're using gpt-5.4, this will consume usage about 30% faster than gpt-5.3-codex.

    • I’ve used gpt-5.4 before, and it’s never overheated like this.

    If you use "fast mode", this will consume usage at a 2x rate.

    • I haven’t used fast mode, so I’m not using twice as much.

    If you use the large context window (a new experimental feature available with gpt-5.4), this will consume usage at a 2x rate.

    • I don’t use a 1M-token context window; I use the default one.

    If you enable the experimental multi_agent feature, subagents will typically consume tokens at a higher rate than if you are not using subagents. As we continue to refine this experimental feature, some changes may affect token consumption. If you are concerned about usage consumption, leave this feature in its default (disabled) state for now.

    • Subagents were used, but there were only a few of them and they were set to “high”... and, more importantly, they were released two days ago, so I assume they're working fine

    Manually enabling feature flags for features that are still under development may consume tokens at a much higher rate. We don't recommend doing this.

    • I haven't enabled any experimental features

    Please check my chat of thread id attached.

  5. YoloMark98 commented on Mar 18, 2026

    @YoloMark98
    Author

    Many people are saying here #14593 (comment) that the issue is always down to CODEX; in other words, too many ‘sessions’ mean that more tokens are being used somehow – perhaps because every agent that is run takes a sort of summary of all the sessions? Is this perhaps why credits are being used up so quickly?

    You absolutely must look into this, as we users cannot afford to lose so much usage over an error that has now almost been identified.

  6. YoloMark98 commented on Mar 18, 2026

    @YoloMark98
    Author
  7. SarthakYadav commented on Mar 18, 2026

    @SarthakYadav

    Please refer to #14593.

    @etraut-openai Before writing off what users are saying, you ought to do a better job of checking into the issues. I just upgraded but still couldn't use my account, even when I could see 100% usage remaining. Deleted my sessions folder and it worked. There is a bug, at least acknowledge it.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    CLIIssues related to the Codex CLIbugSomething isn't workingrate-limitsIssues related to rate limits, quotas, and token usage reporting

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions