Weekly limit reset date changed from Nov 3 to Nov 7th
Open 💬 43 comments Opened Oct 30, 2025 by williamritossa-le
💡 Likely answer: A maintainer (etraut-openai, contributor)
responded on this thread — see the highlighted reply below.
What version of Codex is running?
v0.50.0
What subscription do you have?
Team
Which model were you using?
gpt-5-codex-medium
What platform is your computer?
Darwin 24.3.0 arm64 arm
What issue are you seeing?
Yesterday /status said my weekly limit resets on Nov 3rd.
Today /status says my weekly limit resets on Nov 7th
Today's date is Oct 31st
What steps can reproduce the bug?
Uploaded thread: 019a33d4- 7e1e-7af3-bb6e-c47cce4164e6
What is the expected behavior?
_No response_
Additional information
_No response_
43 Comments
I got the same , not sure if they are extending the limit or something else
I also came here because I've noticed the exact same thing as well. It's a little bit cruel for some of us who are trying to judge how much to use each day and then the goalposts move by five days, but the allowance not adjusting at all - so we have to make our weekly limit last another five days?!
I also came here to report the same issue. I did not touch Codex a few days to get it to reset its weekly limits on monday morning. The limit was supposed to reset on monday 3 november. Now the reset data has been moved to fridays again (also 7 november). This does not give me more usage, it just makes me spread the same usage over a longer period. This is very annoying. I am using claude as well as codex, and i spend my weekly limits in 2.5 days of coding, so i tried to create an offset, between the weekly limits. But like this if feels like open AI took away 4 days of usage for the same price.
me too,Fxxk!
same
me too, very annoying.
It seems that they are doing some sort of rolling window. because mine was again reset but now to 9 november. usage is also back to 0. I wonder if it also works like this if my usage would have been high.
mine now also says nov 9 but usage is %100 although no use at all
Thanks for your post. We're providing updates related to this issue in this thread. I'm going to close this issue to reduce duplication.
Thanks for your reply. But i do not agree that it is a duplicate of that question.
In this topic it is not so much about the reduction of the usage. That there is a weekly limit is one thing, but the weekly limit is moving within the week, so we see resets at 5 nov, 7 nov, 9 nov, 11 nov etc. It seems it is resetting weekly limits in some sort of roling window fasion. Today it happened again. it seems that it is impossible so save usage by for example 3 days not using it and then using everything in 4 days. It seems that if usage is under some limit, the time period is extended.
So this is really a bit different, then the general complaining about lower rate limits. It would just be nice if it were specified what the policy on rate limits is, and when rollover occurs if usage stays below what threshold.
Confirmed it's still happening with 0.56.0: https://github.com/openai/codex/issues/6429
It seems to have been fixed at least i am not having the rolling window, changing reset dates since a few days. Anyone else still experiencing this? EDIT 23.11.2025 No, its not fixed, it happened again. This is so annoying.
My weekly used to reset on Tuesdays, now this week it pushed to Thursday, then changed again to Friday, effectively losing a few days of usage. This is pretty crappy of OpenAI.
same here happened today . I wanted to save up some rate. And did not do much the last 2 days. Today my limit was reset to 100% so I lost about 2 days of coding ~30% of my weekly limit, again! That happened now 3-4 times, really annoying...
This is extremely frustrating. My limit was just reset again today, which I suspect is related to the 'Elevated Errors on Codex' incident on Nov 22.
While I understand you may need to reset limits to compensate for latency or outages, resetting the usage quota should NOT reset the 7-day schedule anchor.
The Problem: A compensation that resets the timer is actually a penalty. Because my 7-day timer was reset along with my usage, my 'weekly reset' date has now been pushed back to today (the weekend), completely disrupting my planning for the upcoming week.
The Fix: Please decouple the Usage Amount from the Reset Date. If you need to clear our usage due to technical faults, leave the 7-day timer fixed on its original cycle. At the very least, send an email notification so we aren't left guessing why our schedules have suddenly shifted.
Hi All, I don't want to hijack this thread, or issue, but a while ago I opened a similar issue at the OpenAI forum. It was less specific but at least there was some response from openai , it might help that besides posting here, if we get a bit louder there, with new issues or posts. Maybe in a more general way, about notifying of resets and explaining that sometimes a reset can be a penalty rather than a compensation, as Xiangz formulated very well.
My Post was: https://community.openai.com/t/weekly-limits-reset-date-suddenly-changed/1364615 But maybe opening another is even better.
happened for me today as well yesterday it was supposed to reset on the 28th (i made sure to immediately run a prompt after the previous weekly limit was reset). Today that was gone and after my first prompt there is a new reset time on the 30th of November. My weekly limit is also back to 100%, but this behaviour is extremely confusing and frustrating.
I can confirm this is still happening
Are people really okay with OpenAI arbitrarily violating ToS and stealing promised credits? I don't see any response from OpenAI team regarding this.
There was a codex service outage this weekend that affected some users in some regions. When a service outage occurs, we typically reset rate limit dates as a way of crediting users for any token usage that may resulted from retries or other side effects of the outage.
Several of you have raised a good point about these resets being disruptive — especially if you plan your codex usage around your weekly reset date. @xiangz19, thanks for proposing a fix. We are considering your idea and other variants.
Is there any plan to provide some kind reprieve for people who were negatively impacted by these resets? I have lost quite a lot of allocated usage because of resets and support has been less than helpful.
My weekly reset was supposed to take place on 28 November at 00:18, but was suddenly postponed to 30 November at 00:03. The reset date is being postponed without explanation. I request that this be corrected as soon as possible, given the price I pay for the Pro subscription! How can something like this happen to a billion-dollar company? It's just annoying! If I have to wait until the 30th, then that was my first and last subscription! Just disable that damn weekly limit until it works properly!
Until you send a message or start a conversation, your weekly rate limit does not get set, even though it has been reset. This is another problem and a classic way of stealing credits!
@etraut-openai Is there anything being done to address this problem for affected users? Support has been acting like they don't understand the issue and is repeating auto generated policy responses straight from ChatGPT.
This just happened to me: https://github.com/openai/codex/issues/7325
My reset was extended, but usage quota was not reset. This will cause me to run out!
Regardless of the outcome, my journey with Codex/OpenAI will come to an end after this period. 229 euros should include a minimum level of service. My post has not been approved on the forum, there has been no response here, and there is no other way to contact a support member. It is all simply a scam. May the AI bubble burst as soon as possible. I'm out of here.
@ferretsparadox, I don't have any update to provide at this time. This is a holiday week in the U.S. (Thanksgiving), and most of the Codex team are taking a well-deserved break so they can spend time with their families. The service is being monitored, but we're not deploying any changes or publishing new clients this week. I'll post again when I have updated information to report on this topic. In the meantime, thanks for your patience.
Mine fixed itself
It's great that the devs can relax a bit, but then this issue must have multiple root causes. My reset time changed just this morning. This time from today 2am (CET) to 1pm.
UPDATE: apparently the logic of the reset time did not change, and only the display changed wrongly. i just ran a prompt at 11:30 am and now my reset time is 11:30 am on the 7th, hope it stays that way.
Weekly limit got 30% depleted by a single task that didn't even finish and consumed my 5-hour limit on a Plus plan. It quasi-looped into fixing pytest errors one by one and retesting, instead of fixing all and retesting. Sometimes the limits go both to 100% and in those "buggy windows" I manage to accomplish something. Otherwise with the current limits and their unreliability and unpredictability, Codex is unusable for any work - small or large, since we can't budget our limits. My reset is on Dec 7, but I don't believe anything anymore.
This is very frustrating — the 7d reset time keeps jumping around. I wrote a script that polls the rate-limit info every 20 minutes; below are excerpts from the recordings:
sample time | type1 | pct1 | reset1 | type2 | pct2 | reset2
-- | -- | -- | -- | -- | -- | --
2025-11-30 16:05:07 | 5h | 38 | 2025-11-30 16:17:57 | 7d | 75 | 2025-12-02 17:00:03
2025-11-30 16:45:04 | 5h | 5 | 2025-11-30 21:18:04 | 7d | 78 | 2025-11-30 16:50:02
2025-11-30 17:05:02 | 5h | 9 | 2025-11-30 21:18:04 | 7d | 1 | 2025-12-07 16:50:09
2025-11-30 21:45:01 | 5h | 0 | 2025-12-01 02:45:01 | 7d | 2 | 2025-12-07 16:50:09
2025-11-30 23:05:01 | 5h | 0 | 2025-12-01 04:05:01 | 7d | 81 | 2025-12-02 17:01:58
2025-11-30 23:25:01 | 5h | 7 | 2025-12-01 04:11:02 | 7d | 83 | 2025-12-02 17:00:03
2025-11-30 23:45:01 | 5h | 15 | 2025-12-01 04:11:03 | 7d | 7 | 2025-12-07 16:50:09
2025-12-01 00:05:01 | 5h | 24 | 2025-12-01 04:11:02 | 7d | 9 | 2025-12-07 16:50:09
2025-12-01 00:25:01 | 5h | 24 | 2025-12-01 04:11:03 | 7d | 88 | 2025-12-02 17:01:58
2025-12-01 09:05:02 | 5h | 0 | 2025-12-01 14:05:01 | 7d | 9 | 2025-12-07 16:50:09
The 7d reset timestamp flips between multiple inconsistent values. This looks like a serious internal data-consistency issue with Codex’s rate-limit reporting.
It happend to me too.
It should reset on Dec 1st. But it shows it'll at 17:28, Dec 2nd !
It seems happens frequently.
It seems codex write the code, and it cannot fix it.
My incorrect reset date turned out to be cosmetic only. It did reset.
Mine keeps jumping around as well, but it seems to like me as its just been refilling empty limits! Codex may be buggy, but sometimes bugs can be nice.
Regarding the problem of the reset date jumping around, here's a quick update. We believe we understand the problem. Earlier today, we took the first step toward mitigating this issue by resetting the limits for everyone in Codex. This was the fastest and least-risky way to eliminate the worst of the inconsistencies that some Codex users were seeing. We plan to roll out additional fixes over the coming days.
Thanks for the update. Probably there was another reset, my weekly reset date moved from 8 to 10 december.
it just happened again for me.
Still happening...
So, when will it reset to 100%?
The rolling window is just a nightmare for me, because it's not just a rolling window with a fixed 7 days.
If my 'rolling window' resets at 7pm Friday, but that week I don't get time to work on something with Codex until Monday at 7pm, then I've effectively 'lost 3 days' - the 'rolling window' does not start until I enter a prompt.
So effectively my 'rolling window' is 10 days in that case. It sucks. I'm tired of adding & deleting calendar events to try and keep track of it now.
I know Codex have to balance load etc, but I wish that the rolling window was just 'the same 7 days', resetting on the same day every week. One situation I recall, is having 40% weekly usage left, planning some tasks the next day to use it, knowing that my Codex usage Dashboard said the weekly usage limit would reset tomorrow at 9pm.....then logging in around 1pm, to find that, contrary to what Codex said yesterday, it had already reset. That wasn't nice.
From reading quite a few threads here on Github and on the interwebs in general, it feels like for a lot of users, this is not 'transparent billing', and you regularly feel like you've had the rug pulled out from under you. I'm sure that's not the intention - but that's what it feels like when it happens....again, and again.
I don't see why we can't just have a 7 day window, based on when you signed up for your plan. Resets every 7 days. Users know where they are, and don't feel mislead. Simples.
I think this happens when the usage is low...?
Recently this has been happening to me again, but like I noted last time, it's just constantly resetting, giving me practically unlimited usage. I'll definitely use it, but it is weird
I had codex for a week, its been two months and my limits didn't reset. The date for reset keeps moving, the weekly reset keeps moving. I was concerned my api keys were exposed so i killed all of them, and still being limited even tho I'm two months into my subscription. Should I cancel and sign up again?
It seems like this is what happens
Let's say you have 7M token allocation for 7 days = 1M tokens per day (can vary by plan etc. but for our simplified discussion, let's stick to this hypothetical scenario)
Scenario1:
If you use 0 tokens on day 1, now remaining allocation is 7M tokens/6 days = average 1.167M tokens per day for next 6 days
Here resetting the window = you again have average 1M per day, so 1M tokens from first day are not "saved", they are lost and not usable by you.
Scenario2:
If you use 2M on first day, now your allocation is 5M/6 days = 0.83M tokens per day
Here resetting the window = you again have average 1M per day, so you got a gift of 1M extra tokens on first day.
The window is not reset because it will have to allocate extra tokens to give you average 1M tokens per day.
While scenario2 is fair, scenario1 is unfair as I am unable to use tokens I paid for. It is a sneaky way to reduce your token usage. But because it is very confusing, the cost of frustrating the customers is much higher than tokens saved.
A better strategy may be to just offer 1M tokens per day (for easier communication and make it less confusing) or do not roll my windows (like Claude Code) so I can plan my workdays and use the quota I paid for.