I haven’t heard anything official, but some Codex Ambassadors have shared that this is not just on the roadmap and will be released soon.
I have a $200 plan, and had my usage reset this morning at 6:40am. It’s now 3:25pm, and I have used 100% of my tokens again… Something has changed in the past few weeks that is causing token exhaustion much faster than ever before, being at 433M tokens. Past WEEKS, I had around 2.6B tokens and not hit the weekly utilization limit. Still… 433M is way better than the 38.2M I got last week.
No sub-agents, no multiple sessions, just 1 active Codex chat, reviewing log files to IDENTIFY bugs within my code that I could then later fix. It didn’t even get a chance to fix them, nor even finish reviewing said log files from the past week. Instead, I went from the 2nd - 6th before being told I had 0% usage remaining…
Hi Codex Team,I am writing to express my deep frustration regarding the recent shift from the 5-hour refresh cycle to a strict weekly limit. Under this new scheme, my tokens drain incredibly fast, often leaving me locked out after just one single day of intense coding.Waiting a full week for a reset makes Codex completely unusable for professional, daily development. As a paying subscriber, I am seriously considering canceling my subscription and migrating to other alternatives (like Cursor or GitHub Copilot) unless a fairer, dynamic, or daily reset option is reintroduced.Please review the community feedback on this matter. A weekly cap is hurting our productivity.Best regards,
Look, I used maybe 3–4 commands and I’ve already hit the weekly limit. This is honestly getting ridiculous — I can’t get any work done like this. I can’t afford to keep paying more and more, and my quota seems to run out faster every time…
FOR THE LOVE OF GOD, PLEASE GIVE US A HIGHER LIMIT!
I am quite sure usage of weekly is not normal. went back to try codex that I pay for 2 seats business - and my weekly limit is zero (0) after just 19 messages. Why would anyone use this ?
- Total lines of code 1948
- CLI 13
- GPT-5.6-sol / Reasoning medium
My quota just reset today—I am a 20Xpro subscriber. However, my weekly quota was fully exhausted in just two hours. Please investigate this issue with the quota. Thank you.
Still going through a full week’s of usage in a few hours without using sub-agents, and being around 488M tokens vs the usual 2-3B tokens for the week…
But hey… at least OpenAI released a new “ultrafast” mode so your tokens and usage can go poof even faster! Their priorities are rock solid and right where they should be.
Has anyone else experienced something similar with Codex weekly usage limits?
I’m on ChatGPT Plus and currently using Codex CLI v0.147.0 with GPT-5.6 Sol.
On August 13, my Codex weekly limit showed about 50% remaining, and the reset date was August 18.
I then stopped using Codex for a few days. From August 13 through August 15, I did not submit any Codex tasks or start any Codex sessions.
When I checked again on August 16, my weekly limit had dropped to around 14%, and then 13%, even though I had not used Codex during those three days.
Even more strangely, the reset date changed from:
August 18 → August 20 at 12:33
My current /status shows:
Weekly limit: 13% left (resets 12:33 on 20 Aug)
I also checked my local Codex rollout logs. My token activity was:
- Aug 10: 15.81K
- Aug 11: 22.88M
- Aug 12: 4.27M
- Aug 13: 0
- Aug 14: 0
- Aug 15: 0
- Aug 16: 1.32M
So according to my local Codex logs, there was zero token activity on Aug 13–15.
I understand that local rollout logs may not represent all server-side or shared agentic usage, so I’m not claiming that this is definitely a bug. However, I find two things particularly confusing:
- Why did my remaining weekly allowance fall from about 50% to 13% while I wasn’t actively using Codex?
- Why did the weekly reset date change from August 18 to August 20?
Could this be related to background activity, memory processing, a weekly-window reset, shared agentic usage, or some issue with rate-limit accounting?
I’ve also contacted OpenAI Support to ask them to check the server-side usage records.
If anyone else has seen their Codex weekly limit decrease while idle, or has had their weekly reset date suddenly move, I’d be interested to know whether you found an explanation.
I’ve spent the last 2 days fixing the most basic UI fixes. Last week it would have been a breeze, today, 40% of weekly consumption was burned, and still can’t solve the most basic changes.
This is very frustrating and disappointing.
In addition, I see this error often:
stream disconnected before completion: failed to send websocket request: IO error: Broken pipe (os error 32)
Same thing here with reset being randomly pushed to aug20. Hopefully another token usage reset is coming, and that is the reason. In itself it’s annoying because it messes up the plans. I’ve learned how to prepare for random token resets, but I am still figuring out how to prepare for random date changes. I really don’t want any more lottery, just stable development.
Have you looked into the API? Pay as you go with bigger rate limits typically.
Prices are ridiculously high, but I am mulling it over.
SAME!!! im getting killed out here
I expected my weekly usage limit to reset on Monday, so I let Codex do some heavy data processing on my server over the weekend, expecting to be ready for work today. But I found that my next reset date had been pushed back to August 20, and I’m now almost out of usage.
I did some searching and found that the Codex team has reset users’ usage limits quite often. By my count, there have been 16 resets since GPT-5.6 was released. That’s great and very generous, but it also seems to push back the next reset date.
Now I have to use Gemini and DeepSeek to finish the rest of my work, which is really annoying. The weekly reset should stay on a consistent schedule, and users should be notified when it changes so they can plan their work better.
I just found many similar topics about this issue here. The team should really take a closer look at it.
That’s why the banked resets were such a great idea, but they aren’t giving those anymore.
I am working with terra on x-high and it really gets shit done.
It is a really reliable workhorse especially when combined with luna subagents.
Roughly 3 million lines of code over 2800 commits in 6.5 weeks is not really nothing.
And that is just on one project - I’ve build a couple websites with it too - and they are not really ai slop bullshit.
I use sol for planing of bigger refactorings or modules like “sol, make a plan to implement xy - write it into our sprint.md and add checkboxes so we can give tasks to luna subagents and have them do the sprint in badges”
Then i switch to terra on x-high and tell it to orchestrate the subagents.
This leads to sometimes hundrets of turns.
Sol is expensive! Don’t waste it!
Did you use chatgpt in work mode in that time? Workmode is fueled from codex limits.
Tell me about it. I’ve purchased additional hardware for the reset and now its just sitting there on dedicated server waiting for reset.