Guys, wake up. Is this a bug or a rate limit decrease?
It’s funny you mention people should wake up while asking whether this is a bug or a rate limit decrease.
There have been multiple update e-mails and update posts about it. It is a rate limit decrease. The 2x rate limit bonus ended per this month.
It’s pretty clear and obvious pretty much everywhere.
This, however, does not change the fact that it is now absolutely unusable. Hitting the 5h limit after 3 prompts is absurd. It’s only working for maybe 12 minutes before you’re hitting the limits. That’s crazy.
Hehe, apparently I was the one who‘s asleep 
Thanks for the answer, although, it is fair to say that my reaction was reasonable considering this was the exact email text that they were informing us:
”(…) We’re reducing the annual price of standard ChatGPT Business seats from $25 to $20, while increasing total weekly Codex usage for users. Usage is now distributed more evenly across the week to support day-to-day workflows rather than concentrated sessions.” (received on April 3, 2026)*
*
Also, I thought they were drastically limiting the usage during the nights etc considering the night hours at my location are peak hours in the US.
Me too!! never ran out of 5h window usage while the 7d limit is going down so quickly, which doesn’t make sense at all. it used to be like you spend a good 10% of the 5h window to get the 7 day 1-2% down, but something’s happened since they gave us that free 7 day window top up a couple of days ago.
my feeling since >2 months: quotas have been drastically nerfed. On Claude and gemini too.
So I tried to deepseek v4 flash, to do the same tasks. It did a lot more and faster, for free. If this quotas thing is still here in few weeks I will definitely move. Sorry openai, competition is rude.
I’ve written about this before, but here we go again — Codex usage limits run out faster than expected. I tried optimizing prompts: writing concisely without extra whitespace or line breaks, asking for conclusions only, no explanations — straight to the point. That actually works and can save a chunk of your quota, worth keeping in mind. I also tried dropping down to model 5.2, since the tasks were narrow — no complex problem-solving, just replacing specific code in specific lines. The difference wasn’t big: with model 5.5 the quota was gone in about 40 minutes, with 5.2 around an hour. In my case the bulk of the quota goes to frequent code checks, tests, project audits, and logging work maps — that’s unavoidable. The same old issue with limit blocking is still there too: when the weekly limit runs out early, access is blocked even though tokens still show on the balance. You simply can’t use them. I’m not alone — I’ve seen similar posts on the forum. The company’s response was to reset limits and hand out a couple of bonus hours. Meanwhile, any unused balance just rolls back into OpenAI’s pocket next month.
The thing that bothers me most is that when selling the plan, the important data is not transparent.
It only says PRO 5x more than PLUS and 20x more than Plus.
then I tried the 5x and the 20x and the difference between them does not equals 4x
In the Codex IDE, showing 5h and Week quotas as % is also misleading.
Is it a request? A token? What? Then it should show a 5 h window, 200 requests left, and so on not as a percentage.
Because I see the model working and the percentage dropping, but that does not allow me to pinpoint what it is consuming. so I can finetune my work to save usage.
I think I should also complain about the amount of quota, which is either too low or the model uses too much.
I am BURNING through my allocation, even if I just use 5.3-codex. It feels like if you touch 5.5 at all, your whole ability to get things done will just go up in a puff of smoke.
5.5 is great when it’s actually doing stuff. I just run out of the ability to do anything very very quickly.
I hear you, KevinR. As you can see in my screenshot, I finished my week a few days ago. I still have a long way to go before I hit the sanity limit. I can see how that motivates everyone else to do something 
