Rate Limits dropped 30% within a few hours of use? Teams account - not getting usage resets and still getting fast drops
What issue are you seeing?
Use account ID: user-0I6wEVvjWlQ8g0LXj6oYl0KN, user-AC8w7BaA2CLBYd76mO59bV3A (since I have a Team account - these are the two IDs).
Have been using GPT-5.4 on high - and I am aware of the 20% faster usage. However I have been using it for a while - and today I see usage dropping extremely fast. This is different from my everyday experience - I've been using codex for a while. I use opencode with the OpenAI codex sub.
Usage on Account 1
<img width="1283" height="1102" alt="Image" src="https://github.com/user-attachments/assets/e380712a-3d37-45f5-88cb-90361f870573" />
Usage on Account 2
<img width="1275" height="1133" alt="Image" src="https://github.com/user-attachments/assets/33136fd0-5f06-4330-bf15-aefc2b71662c" />
Dont use fast mode, opencode defaults on GPT 5.4 high.
Teams user - have not been getting all of the usage limits reset (maybe that is because I wasnt affected but looks like all Plus users got usages reset? - even who weren't affected, but Teams not being fixed and usages reset is really annoying.
What steps can reproduce the bug?
Using it as usual - gpt-5.4 high, usage dropping really fast.
What is the expected behavior?
_No response_
Additional information
_No response_
13 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
Not a duplicate as X posts say this issue has been fixed.
This looks like a duplicate of my issue: https://github.com/openai/codex/issues/14329
But yes, perhaps they need as many complaints as possible in order to take this matter seriously.
We're tracking this in #13568.
I understand -- however the resets aren't being applied to Teams and thus we've been kind of losing out for a bit here. Also OpenAI status doesn't show this as an outage anymore - so I decided to file a new issue.
@etraut-openai 2nd that RE: Teams/Business Subscriptions Checking for progress on this and #13568. I just spent >$100 in credits (within the last 3 days) and they disappeared in 2 days, over a few hours of basic code use. Am I able to get a refund for the credits until the issue is solved???? I'd forgotten this was still happening. Then I couldn't even use the thing well because you guys had an outage the last two days. I would definitely like more credits or some of my spend back.
I've been finding this exact same thing. My account reset on Monday, and I've never managed to hit my weekly limits before, but today I'm now down to 3% of my weekly limit with no change of usage except moving to GPT-5.4. I opened a support case (06615214) to ask them to investigate. Also on Tuesday I had hundreds of reconnecting messages, but I didn't think to check the usage then. But something is definitely wrong because there's no way that I've used that much Codex in only two days, with no change of usage - in fact, probably slightly less as I've been concentrating on one project, rather than running two at a time like I often do.
Edit: I don't actually know that my account reset on Monday as it was supposed to. I didn't check it until today, so it may well have not, which would explain why I am now so far through my weekly limit.
Edit2: My quota has now dropped to 0% and I'm locked out. If I buy credits to continue will we be able to get them refunded when you sort the issue out?
Is there a quick link to reset the limit till this issue is resolved? My workflow is struck
@etraut-openai if this is being tracked in a single issue - could we atleast have updates in there?
@etraut-openai
https://github.com/openai/codex/issues/13568#issuecomment-4050697702
I’m replying here because you’ve acted like a good politician who doesn’t allow any response to his statements. The bottom line is that, almost two weeks later, you still haven’t solved the main problem, which was the high usage/consumption.
As you know perfectly well, because you yourself closed issues, this has been happening since before the 5.4 release, so it makes no sense to blame that version. You have countless reports, including mine, explaining it and providing figures and charts showing that this was happening without enabling 2x or using multi-agent setups.
Basically, according to you, it’s our fault — hundreds or thousands of us supposedly coordinated to change the way we use Codex, and all at the same time.
Fortunately, we postponed hiring/purchasing more accounts until this problem is solved, which, as we suspected after so many days, is not going to be fixed. After all, you seem quite comfortable with the increased usage/consumption.
@etraut-openai
I'm experiencing the exact same issue, and there's no way this is acceptable — even before factoring in the 2x multiplier. We should never be having this conversation at 2x to begin with, yet consumption is so extreme that usage is completely drained within a single day.
If this is considered normal, developers will be cancelling their subscriptions en masse come April 11th. Claude Code gives significantly more tokens with Opus than Codex does at this rate — and we never had these problems with Codex ever before, Codex 5.4 was supposed to be more token-efficient, and OpenAI is racing to match Claude Code on speed — so how is it possible that this many developers are reporting the exact same problem while OpenAI insists everything is fine?
I really hope they will put more effort into finding the issue. otherwise we will need 3 subscriptions per week.
It's not fine.
Because they clearly have a problem and don't know where to look or they've started nerfing usage under the guise of that there is nothing wrong.
rate limits today seem to be dropping 2-3x faster