Repository navigation
Codex usage limit consumed ~10x faster (1 message ≈ 6% of 5h window) — likely bug #16889
Closed as duplicate
Description
What version of the Codex App are you using (From “About Codex” dialog)?
Latest (desktop + CLI, issue started ~3 days ago — didn’t happen before)
What subscription do you have?
Plus & Business
What platform is your computer?
macOS (MacBook, Apple Silicon)
What issue are you seeing?
Since ~3 days ago, Codex usage limits are being consumed dramatically faster than before (~10x increase).
Even a single normal message (no large code, no heavy context) consumes ~6% of my entire 5-hour window. Previously, similar usage would consume only a small fraction of that.
This happens both in:
• Codex desktop app
• Codex CLI
This makes Codex effectively unusable for normal workflows.
What steps can reproduce the bug?
1. Open Codex (desktop or CLI)
2. Start a new session
3. Send a normal message (no large code or complex context)
4. Observe usage consumption
Expected:
• Minimal usage consumption for small/simple messages
Actual:
• ~5–6% of total 5-hour limit consumed per message
What is the expected behavior?
No response
Additional information
No response
Activity
github-actions commentedon Apr 6, 2026
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
ayaryabi commentedon Apr 6, 2026
Hey @etraut-openai ,
My issue was closed as a duplicate of #14593 — a thread that's now 3 weeks old with no resolution. I
understand deduplication keeps things organized, but closing new reports without investigation isn't the same
as fixing the underlying problem.
A few things worth noting:
more reports now than when Burning tokens very fast #14593 was opened. Something appears to have changed again recently, possibly
around this past Friday.
look contained when it isn't. The people commenting "not a duplicate" in the main thread are saying exactly
this.
that this is a known regression. Closing new issues into an unresolved parent doesn't help paying users.
I'm not looking for a workaround. I'd like to know: is this actively being investigated? Has anything changed
on the backend in the last few days that could explain a new wave of reports?
oerx commentedon Apr 7, 2026
Currently, you get more with the Claude plan than with the OpenAI Business plan. If your complaints are being ignored, the best feedback is simply to switch providers
thanosd-bbw commentedon Apr 9, 2026
Thinking the same as you both.
Patrick-Glatt-Vitra commentedon Apr 10, 2026
Same here...
After ~10 prompts the first 5h are over. That's insane!
Our Company is interested in AI stuff but this is a perfect example why they should stay independent.... This issue exists over weeks and nothing is fixed.
Looks more like a hidden change on price model, pilot project for new prices whatever...
Edit: even worse today. its ~ 5 bigger promts and the 5h are gone
eumarumar commentedon Apr 10, 2026
I keep hitting same problem since yesterday. I hope it is resolved as soon as possible
sontx commentedon Apr 20, 2026
For me, it's only 1 prompt 🤨 what happens to openai, I'm switching to github copilot for now
Tony-Emad commentedon Apr 20, 2026
Same with me after one message my usage consumed and this issue is from one week and my ticket is closed as duplicated kindly check @etraut-openai
soukainaelme commentedon Apr 27, 2026
Same for me. I only used it for a single command—I was at 100% of the 5-hour limit, and now I’m at 0%. I don’t understand. The command was just adding a log, something very basic that shouldn’t require much processing. I really don’t get how this is possible. I already saw that last week but today it was worse
sabrawy-sys commentedon May 8, 2026
last month with this nonsense no moew codex