Limits decrease without using any OpenAI product

Open 💬 2 comments Opened Mar 18, 2026 by nishikawa7863
💡 Likely answer: A maintainer (github-actions[bot], contributor) responded on this thread — see the highlighted reply below.

What issue are you seeing?

Like many others affected by the aggressive weekly limit consumption, I completely stopped using all OpenAI products and switched to an alternative. My weekly quota resets on Wednesdays. I logged back in today (Wednesday, March 18) to check my account status, expecting a full reset. Instead, I'm already showing 3% consumed on the 5-hour rolling limit and 1% consumed on the weekly limit, without having opened a single OpenAI product since last week.

<img width="1286" height="632" alt="Image" src="https://github.com/user-attachments/assets/c910d026-b91f-4f6f-ae62-3f34cd75da52" />

What steps can reproduce the bug?

Stop using all OpenAI products entirely (CLI, ChatGPT, API. everything).
Wait for the weekly quota to reset (Wednesday).
Log in and open the Codex CLI.
Check the usage/limits dashboard.
Observe that usage has already been deducted despite zero activity.

What is the expected behavior?

After a full reset cycle with zero usage of any OpenAI product, both the 5-hour rolling limit and the weekly limit should show 0% consumed / 100% remaining.

Additional information

  • Account type: Pro
  • Last week my weekly quota was fully exhausted by Thursday despite moderate usage.
  • I have not used the CLI, ChatGPT, the API, or any OpenAI-integrated tool since before the reset.
  • No background processes, IDE integrations, or third-party apps are configured to use my OpenAI credentials.
  • Screenshot of the usage dashboard showing the phantom consumption is attached.

View original on GitHub ↗

2 Comments

github-actions[bot] contributor · 5 months ago

Potential duplicates detected. Please review them and close your issue if it is a duplicate.

  • #14023
  • #13330
  • #13975
  • #15039

Powered by Codex Action

DeanStr · 5 months ago

Possibly similar to what I've observed, it was just more visible because it was coming out of my spark usage https://github.com/openai/codex/issues/13854