Issue with GPT-5 Codex Limits Despite Having Pro Subscription

I’ve canceled my subscription for now. If they fix the limits, I’ll come back. Too bad, because Codex was literally life changing to me.

I will be watching before my sub renewal to see, hopefully they listen to their customers. Currently downloading Trae on my VPN, hoping this works as Canada is not supported. If Trae does what I need it to, I’ll be happy to ditch OpenAI.

I had one 6 minute task that updated 60 lines and fails tests.

That’s 52% of my 5 hour limit on Plus.

I thought Claude Code usage was bad but now it looks AMAZING compared to the joke Codex has suddenly become.

I highly recommend trying TRAE, I just tried this based off someone elses suggestion in another thread and it blows the newly gimped Codex out of the water.

Are you using TRAE Solo or the IDE? I’m definitely going to take a look, and it seems cheap compared with ChatGPT now that Codex is worthless.

I’m facing the same issue as people described here starting from today. My quota was empty after two tasks and will be reset at 10/31 02:00AM, after one hour without use codex, my quota is now 99% and the entire day my quota usage was very crazy, I hope this is a bug.

I’m trying with the IDE and it has been great so far, and to try for the current $3 for the first month was a steal. I already cancelled my Plus sub, OpenAI has a few weeks until I decide if I renew again.

This is not a bug, this is OpenAI’s attempt at trying to get as much money as possible from their tools.

Ironically if their usage limits were reasonable the $200 plan could be worthwhile - but at these usage limits even that would be just enough to trash a project and leave it unfixable.

I’m really sick of OpenAI springing crap on its users so unless they fully reverse course (which is surely unprofitable) I’m out.

Really appreciate the tip about TRAE!

Thanks everyone for surfacing this. A recent change in how usage was calculated ended up counting tool output tokens, which caused quotas to drain way faster than expected. This wasn’t an intended change.

The team is rolling that back and reviewing refunds for anyone affected, and I’ll post again once that’s live. Thanks everyone for calling this out.

Keep us posted, i just woke up and can’t work anymore due to a small prompt in the Pro subscription

Still waiting for OpenAI to fix the quota issues because I’m still stuck with a drained weekly quota. I think they just messed up and are claiming it’s a bug to be honest – no smoke tests were done to ensure there wasn’t a severe token bug, do they have a QA that even ran tests? It seems something this critical would have been easy to find.

They probably were testing the waters with this new token based system. To go from using unlimited from release, to gimped in one day, they didn’t think anyone would notice?

Been using TRAE for 48 hours and it’s been a great alternative.

I haven’t had any quota restored yet, but token usage is reasonable now for me (Pro plan currently). I haven’t been able to use anywhere near the 5 hour quota so far.

Keep us posted with the updates.

My weekly quota has been restored! Thanks OpenAI Support for making it right.

Subject: Hit 5-hour limit in just 2 hours on Pro — and already at 35% of weekly cap

Hi team,

I use Codex Web. After usage limits were introduced, I upgraded to Pro.
Today, after about 2 hours of work, I suddenly hit the 5-hour limit and I’m already at 35% of the weekly limit.

What’s confusing is that I’ve reduced my workload to stay within limits:

  • I stopped running 4 parallel jobs; now I keep things mostly single-threaded.

  • Overall task volume is roughly half of what it used to be.

This feels unfair, especially because Codex doesn’t produce perfect code and I often have to iterate and rerun tasks to fix issues — which naturally increases usage even for the same feature.

Could you please clarify:

  1. How Codex Web usage is measured/charged (per request, tokens, wall-clock time, retries, background work)?

  2. Whether there’s a bug or recent change in accounting for Pro users.

  3. If there’s a way to raise limits or get more transparent real-time usage breakdowns so we can plan work responsibly.

Thanks!

Can you guys do something about this limit, it’s so annoying everything weas working nicely but on October 30th 2025 the limits got like so strict even on a paid pro account, i asked the chatgpt codex to do a few tasks and i was at 99% usage limit at the time and the tasks was barely a few hundred lines worth and then i went to go ask a follow up and evidently it says “Usage limit reached” and I came from a plus subscription because I thought this was an issue with plus only and upgrading did not solve it the usage limits are still as strict as ever, now I regret wasting $220 down the drain for this.

Quota calculation tweaks could still be in progress. Last night, I was only able to use less than 10-20% (IIRC) of 5-hour quota. Today, I was able to exhaust it.

For example:

Started with: 12% remaining of 5-hour quota; 74% weekly quota.
Refactored a 4,379 line Python script.
Took 46 minutes
Ended with: 1% of 5-hour quota; 70% of weekly quota.

5-hour quota supposed to reset at 1:14 AM and it’s 1:27AM with no reset yet. Maybe some daylight savings time glitch?

[Edit] it reset the second time 1:14am came around.

This happened again can we expect it to be reset again? It was working normally for last day and now it’s happening again. A few prompts and your through your daily limit.