Repository navigation
Quotas draining like crazy: burnt 100% of my 5h limit in 10 min on Business #20153
Description
What version of the Codex App are you using (From “About Codex” dialog)?
v0.125.0
What subscription do you have?
Business
What platform is your computer?
macOS
What issue are you seeing?
Hi,
I've been using Codex since day one, but I now feel very disappointed.
I burned through my 5-hour quota in less than 10 minutes this morning using 5.5 High in plan mode and 5.5 Low for implementation.
About a month ago, it felt almost unlimited, but now I’m getting a Claude-like experience!
Is there a bug causing this?
Thanks
What steps can reproduce the bug?
Just use Codex
What is the expected behavior?
No response
Additional information
No response
Activity
github-actions commentedon Apr 29, 2026
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
mnj commentedon Apr 29, 2026
Yep, had to switch back to 5.4, which is working as expected with much lower drain on credits, if the usage limits for 5.5 is working as intended, then it's barely useable at all on a business plan, for anything but two simple prompts for minor code changes every 5h.
But "luckily" 5.4 is pretty decent even compared to 5.5 for the tasks we use it for here at work, didn't notice any real code quality benefit with 5.5 for our codebase over 5.4
daletman commentedon Apr 29, 2026
Had exactly the same experience today. One prompt with 5.5 xHigh running for 10-20 minutes and quota was 0. That's just useless.
indyavik commentedon May 27, 2026
I am seeing this too. specially on 5.4 quota is disappearing very fast
t-var-s commentedon Sep 1, 2026
I’m seeing what appears to be the same issue on a ChatGPT Business seat, and I have a fairly clear reproduction from today (Sep 1).
In a single existing thread:
So one prompt consumed 84 percentage points of the 5-hour quota in about 10.5 minutes.
This has not been a one-off for me. I’ve seen repeated cases with different projects where a single thread/new prompt consumes roughly 80% of the 5-hour allowance within around 10 minutes.
I understand that High reasoning and a large existing context can consume more tokens, but this level of depletion seems disproportionate, especially with Fast mode disabled and only one foreground request.
This looks very similar to the original report in this issue, except I’m now reproducing it with GPT-5.6 Sol.
t-var-s commentedon Sep 2, 2026
Sent feedback ID 01a05e14-a9d5-7962-a1b4-6eb9d3cfb643 for this specific thread.
The full 5h usage window started by resuming a previous thread that was interrupted while 5.6 Sol High tried to debug why these usage spikes keep happening. Context window was more than half full, so I called /compact and the 5h usage went from 100% to 94% before I asked the agent to resume its work (which you have to do now since threads get interrupted when usage hits 0).
Agent responds, some change to AGENTS.md is suggested, 5h usage goes to 93%. I close the ChatGPT desktop app (version 26.825.51511 on Intel MacOS 15.7.7) and open it again to make sure the change gets applied. Start a new thread and ask for a plan. That's when usage goes from 93% to 9% in ten minutes.
In summary:
resume previous thread with a /compact - 94%
ask to finish work that was interrupted by the 5h limit - 93%
restart app, new thread, prompt for a plan - 9%
I have similar cases to this from the previous week and the week before.
Edit: Following morning (Sep 2), new thread, implement that plan, usage gone again already.