I have an active ChatGPT Plus (Pro) subscription, but I am facing a serious issue when working with the GPT-5 Codex model in Codex IDE.
After just a couple of hours of active use, I get a message saying that I’ve hit my usage limit and can’t continue working with the model. At the same time, the system suggests “upgrading to Pro,” even though I already have an active paid Pro subscription.
As a result:
I cannot fully use the service, even though I subscribed specifically for this purpose.
My project-related tasks remain unfinished.
There is frustration with these restrictions and confusion about OpenAI’s policy: why are the limits so heavily restricted for Pro users?
I would like to get an official clarification on:
Why doesn’t the Pro subscription allow me to properly use GPT-5 Codex?
What exact limits are applied to this model?
Are there any plans to adjust or expand the limits for users with an active subscription?
This issue significantly impacts my workflow and project development. I would appreciate your response.
I added the official Codex extension for Cursor. I make around 10 requests in a couple of hours, since the models take a lot of time for reasoning. This is not a problem if the model is moving toward solving the issues, but the usage limits prevent me from completing tasks and reaching the correct solutions.
Speaking of limitations, it’s just a nightmare, that is, I literally won’t be able to use the Codex for 3 whole days, I don’t understand why I’m even paying money if I can’t fully use this technology
From that error message you are getting, it seems you have not linked your cursor to your Pro account, if you notice the “upgrade to Pro message”
There should be a be a link account option you can select. Personally, I use the IDE extension in VSCode, so I’m not sure of the exact option in cursor, but I assume it’s a similar option under the CODEX window chat box.
i got this message “You’ve hit your usage limit. To get more access now, send a request to your admin or try again later.”
using team plan. (workspace) with codex.
I had to log out and login again and it started working.
There must have been a silent update to how Codex tokens works as of today, I have been using this no problem since release and today mysteriously I am now hitting usage limit reached and I didn’t even have that many tasks today. Time to start looking for an alternative to Codex I guess.
I ran out of quota earlier today on the “Plus” plan. So I bought the Pro plan and 32 requests later I’m out of quota again. I reckon I could buy 10 Plus subscriptions for the same price and get more quota than one Pro subscription.
It’s one thing to change usage limits but it’s another to do it with no warning, day-of, and then mangle the backend logic so you’re out of credits, no wait you’re not, wait yes you are…
Today has been a great day for Claude Code in my mind. I’m getting really sick of the way OpenAI does things.
They keep proving their incompetence time and time again. This is a no-brainer, if they don’t fix it soon, it’s an easy cancelled sub from me – an alternative should be easier to find with this now gimped version of Codex.
I tested two tasks, one was just to create a task and the other was a refactor of about 100 lines of code and it went from 100% to 65%. I will be cancelling tomorrow and hoping to find an alternative quick.
In 2 hours I used 42% of my weekly limit, and while I was typing this your message appeared and scared the crap out of me because it’s likely what I’ll be seeing myself after one or two tasks.
It’s obvious that OpenAI’s motivation for this was money, but they gimped it so much that anyone using it previously will notice this vast difference. They basically killed Codex – and showed how greedy they are in the same action, not a smart move.
For the next 5 hour block on Pro plan, I’ve already hit my limit after “taking it easy”. My weekly limit has 40% remaining and resets Nov 6th. There is either something really wrong with how they are calculating quotas, or they have indeed killed Codex.
Yes, unless they pivot and fix this, they have killed Codex. This is not surprising behaviour from OpenAI, they have done this numerous times with providing silent updates that either break or severely hinder usage, and this was the biggest one they have done so far. While showing how greedy they are at the same time.