The title is pretty explanatory. I have ChatGPT Plus, and Codex is really a good partner helping me with my project. It’s not like I can’t do without him, but with him, it’s a bit easier to keep going. Help is appreciated, thank you.
SS Attached.
Как это исправить?
Welcome to the forum!
When you run out of your weekly that’s all that is left on that account for the week my friend.
So what about my 5h Limit?
they are tied to the same limit.
When weekly is zero you are on empty.
I genuinely don’t understand this.
I hit the weekly limit, but I STILL had my 5h limit left… which then became completely unusable because the weekly limit overrides it anyway.
So what exactly is the point of showing users they still have 5h remaining if they can’t actually use it? That just makes the limits confusing and honestly misleading.
I bought ChatGPT Plus because it was supposed to give more usable access and fewer restrictions, especially for longer project work. Instead it feels like there are multiple hidden walls stacked on top of each other.
I’m not even complaining about limits existing — I understand resources aren’t infinite. But if the weekly limit completely disables the 5h limit, then the 5h limit practically means nothing once weekly hits zero.
That’s the frustrating pert.
And honestly, the most frustrating part is the confusion.
If the weekly limit is the actual final limit, then why does the app still show I have 5h usage left? Anyone would naturally think they can still use it.
It just feels bad seeing “you still have usage remaining” while being completely locked out at the same time. That’s what makes this whole thing frustrating as a paying user.
Most of us bought Plus so we could actually rely on ChatGPT for projects, coding, studying, work, etc withuot constantly worrying about limits every second. But right now it feels like there’s always another wall waiting behind the first wall.
I’m not asking for unlimited usage or anything crazy. I just think the limits should be way clearer, because right now it honestly feels misleading.
I’m honestly just as frustrated, and I’m having a very similar experience. The issue is not simply that limits exist. The real problem is that the product shows “5h” as if it means usable working time, but in real work on a serious repository that allowance can disappear in about an hour or after only a few tasks. For a normal customer, “5 hours” sounds like 5 hours of available work, not a hidden token / compute / internal usage budget. I already wrote that I consider this limit presentation misleading. If the allowance depends on repository size, context size, model choice, task complexity, tool usage, and token consumption, then the interface should make that very clear. Right now, “5h remaining” creates an expectation that does not match the actual experience. I’m not asking for unlimited access to the strongest model. What I’m asking for is continued basic access for paying users. For example, after the top-model allowance is exhausted, users should still be able to continue with a lower model for audits, small tasks, tests, routine checks, or minor edits. Right now Codex can burn through the allowance quickly and then become basically unusable for real project work. It is even more frustrating if Codex consumes usage while reading project context: repository instructions, files, system context, tool context, and other required information before the user even receives useful work. If that startup/context-reading overhead counts against the same limit, then the system is even less transparent and less fair. Users need a clear usage breakdown: how much was spent on reading context, how much on actual coding, how much on tool execution, and how much on model reasoning. Without that transparency, the “5 hours” limit is not meaningful. It feels misleading for paying users.
I had an interesting conversation with Claude because, after the Codex situation, I started seriously thinking about where to move and which plan to choose. With Claude, the structure looks simpler: less pricing confusion and a clearer access model. Naturally, the main thing I wanted to understand was the limit system. Claude was trying to reassure me that the limits would not work like Codex, with hard weekly lockouts, and that access would be more like daily usage with pauses and recovery. But one part was honestly amusing. Claude itself explained that they had already received complaints because of a huge influx of users and not enough resources to support that demand. Then it basically tried to push me toward the $100 plan. At that moment, I realized that Codex seems to be using a very similar pattern: strong marketing first, a massive wave of users, then strict limits, and then a soft push toward more expensive subscriptions.
From the user’s side, it feels like this: if you pay less, you get less, even if the product was originally marketed as a serious working tool. Maybe OpenAI attracted too many people too quickly and now cannot properly support that demand. It looks very similar to what happened with Claude.Right now Codex has had a huge marketing boost. Many YouTubers were saying how great the 5.5 model is, and after frustration with Claude limits, many users moved toward Codex. But now the boat seems to be tilting in the other direction. My conclusion is simple: the problem is not only the existence of limits. The real problem is how companies sell the expectation of a serious work tool, and then the actual experience becomes a constant fight with restrictions. Paying users should clearly understand how much real work they can do before they pay, not discover it only after hitting the wall.
Yeah, this is exactly the issue. Most people are not even angry that limits exist — we understand these models are expensive to run. The frustrating part is that the way the limits are shown creates expectations that don’t match the actual experience.
When someone sees “5h remaining,” they naturally think they still have usable time left. Not “you might get locked out after a few heavy tasks because there’s also another hidden weekly cap underneath.”
For people using Codex seriously on real repositories, usage disappears insanely fast because the model is reading files, instructions, context, tools, reasoning, etc before it even starts doing actual work. Then suddenly you hit a wall and the entire workflow dies.
That’s what makes it feel bad as a paying user. It stops feeling like a reliable work tool and starts feeling like you’re constantly trying to guess invisible limits.
I honestly think most users would be way less frustrated if the system was simply more transparent about what is actually being consumed and what the real usable limits are.
Honestly this whole thing feels extreemly scummy.
Don’t advertise “5h remaining” to paying users if the reality is that another hidden weekly wall can completely disable it anyway. That is exactly why people feel misled right now.
I paid for Plus specifically to avoid constantly running into limits while working on real projects, and now I’m sitting here with “5h remaining” on my screen while the product is basically unusable because the weekly limit silently overrides everything. How is that not frustrating?
And the worst part is how fast the usage disappears during actual repository work. Codex spends usage reading instructions, files, context, reasoning, tools, and setup before you even get meaningful output back. Then suddenly your entire week is gone after a handful of serious tasks.
At that point the “5h” number stops meaning anything. It just feels like marketing wording instead of a real usable limit.
I’m not asking for infinite free compute. I’m asking for honesty and transparency toward paying users. Right now it genuinely feels like people were attracted with the promise of a serious AI work tool, only to discover layers of restrictions after already paying.
That’s the part that leaves a really bad taste in people’s mouths.
This is honestly exactly the point people are trying to explain, and I feel like a lot of the frustration is being misunderstood.
Most of us are not angry just because limits exist. We understand these models are expensive to run and obviously there has to be some kind of restriction system. The real issue is the way the limits are presented versus how they actually behave in real usage.
When a paying user sees “5h remaining,” the natural assumption is that they still have around 5 hours of usable working time left. That is what almost anyone would think. But in reality, on a serious repository, that usage can disappear incredibly fast because Codex is consuming usage on much more than just the visible replies it gives you.
Before you even get meaningful work done, the system is already reading repository instructions, scanning files, processing context, loading tool information, reasoning through dependencies, understanding the project structure, and handling all the background work needed to function properly. And from the user perspective, all of that seems to count against the exact same limit.
So what ends up happning is:
you sit down thinking you still have plenty of usage left, then after a handful of real tasks, suddenly the weekly limit is gone and the entire workflow just dies.
That is the part that feels frustrating and honestly misleading.
Because if the real limitation is based on compute usage, repository size, context size, reasoning depth, tool execution, and model complexty, then the interface shuld explain that clearly instead of making it look like simple time-based access. Right now the wording creates expecttions that absolutely do not match the actual experince for many users.
And for people using Codex seriously for development work, this becomes mentally exhauting very quickly. You stop focusing fully on your project because part of your brain is constantly worrying:
“Is this task going to burn the rest of my limit?”
“Is opening this repository wasting usage?”
“Should I avoid asking follow-up questions because the context is too large now?”
That completely ruins the feeling of Codex being a reliable profesional tool.
I genuinely think a huge amount of this frustration would disappear if there was simply more transparency. Show users what actually consumed the allowance:
-
context reading
-
repository scanning
-
reasoning
-
coding generation
-
tool execution
-
model usage
At least then people would understand where the usage went.
Because right now, seeing “5h remaining” while getting locked out shortly after makes the limit feel meanigless from a user perspective, especially for paying subscribers trying to use the product seriously
My opinion is that they created such a confusing maze that they probably didn’t expect this result themselves. Maybe they asked ChatGPT to write a prompt, and it came up with something like this. 
Jokes aside, the feeling is like buying a bicycle and then, an hour later, someone takes away the seat. 
I’ve never actually had that experience with a bicycle, but working with Codex gives me a very similar unpleasant feeling.
That bicycle example is honstly perfect 
It really does feel like:
“Congrats on buying Plus, here’s your better experience!”
Then the moment you actually start seriously using Codex on real work:
-
weekly wall hit
-
5h suddenly meaningless
-
usage disppearing while context loads
-
workflow dead
The frustrating part is that the product itself is genuinely amazing. That’s WHY people are upset. People want to rely on it for real development work, but the current limit system makes the experience feel unpredctable and stressful instead of reliable.
Nobody likes sitting there wondeing if opening one more file or asking one more follow-up question is about to vaporize the rest of their weak.
*Most new and existing customers across all plans have been migrated to the new token-based pricing, and should use the new rate card listed in this page. *
A small subset of Enterprise customers should continue using the legacy rate card until we migrate you to the new token-based pricing for Codex. For more information contact OpenAI sales.
>https://help.openai.com/en/articles/20001106-codex-rate-card#legacy-rate-card
Thanks! now how about you let the small guys that are not using this tool for corporate bs to use the old rate card?
At this point I genuinely think OpenAI needs to reconsider how this is presented to paying users, because right now it feels extrmely misleading.
If the weekly limit completely overrides the “5h remaining” limit, then showing users they still have 5 hours left makes no sense and only creates frustration. A normal person sees “5h remaining” and assumes they can still work. Instead, you hit a hidden wall and the product becomes basically unusable for days.
And honestly, being locked out for almost 4 days after heavy project usage feels terrible when you’re paying specifically to use this as a serious work tool. Especially when a huge amount of the usage appears to be consumed just from context reading, repository scanning, instructions, and setup overhead before meaningful work even begins.
I’m not asking for unlimited free compute. I’m asking for transparency, fairer fallback access, and limits that actually match what is being shown to users.
At the very least, I seriously hope OpenAI considers resetting affected users’ weekly limits or offering temporary relief while this system is being figured out, because right now a lot of paying users feel blindsided and disappointed by how this works in practice.
@OpenAI_Support
That’s honestly the part that stings the most.
So, Enterprise customers and large companies get to keep the older, more usable system for longer, while normal Plus users, students, hobby developers, indie developers, and smaller creators get thrown straight into the stricter token-based system?
That feels backwards.
The people most affected by these aggressive limits are the smaller users who can’t casually jump to $100+ plans every month just to continue working normally. A lot of us are using Codex for learning, personal projects, open-source work, side projects, studying, or trying to build something real without corporate-level budgets behind us.
And the frustrating thing is that the product itself is genuinely incredible. That’s why people are upset. People WANT to use it. But the current experience makes many paying users feel like they’re being punished for actually using the tool heavily.
At the very least, there should be some kind of fallback mode after the high-end allowance runs out: Slower responses, Lower priority, Weaker model, Reduced tool access, and daily cooldowns
Anything is better than:
“Weekly limit hit. Come back in 4 days.”
Because for real project work, that basically kills momentum completely. Whoever came up with this, I hate you, man. I hope both sides of your pillow stay warm for the rest of your life.
EDIT:
I can accept limits existing. I can accept pricing changes. I can even accept stricter quotas for heavy users.
What I can’t accept is a UI that says “5h remaining” while the actual product response is:
“Weekly limit reached. Come back next week.”
That’s not transparency. That’s contradictory UX, and honestly it feels like this flow was pushed out without anyone seriously checking whether the messaging even made sense from a user perspetive. If the UI says “5h remaining” and the real answer is “locked out for 4 days,” then what exactly was the point of showing the timer at all?
And now I can’t even properly reply in this thread because apprently new users get rate-limited on responses too, so I have to edit an existing comment just to add clarification. Which honestly makes this whole experience even more absurd.
The issue was never “give me infinite usage.”
The issue is that many of us are using Codex for real project momentum, debugging sessions, architecture planning, and long development flows. A hard weekly shutdown mid-workflow doesn’t just reduce usage — it destroys continuity.
And judging by this thread, I’m clearly not the only person confused by it.
Anyway, I said what I needed to say. Hope the feedback reaches the people designing this stuff.
You need to come to terms with this reality and look for another option.
I’m a student and this is tough. I should just use free version. Thank you openai. god.
I hear you, friend. Every person who buys a product or service wants to know exactly what they’re getting. The problem is that the limit tracking system is either still in beta itself — or it’s some kind of joke. When you buy a physical product you can touch it, measure it, evaluate it visually. Nobody buys a broken bicycle knowing it’s broken. But when you’re sold a service — it’s wrapped in vague terms, conditions and rules where the actual experience rarely matches what was promised. And I’ve noticed something interesting: the forum admins carefully step around these complaints. No official statement, no “we’re working on it”, no “we’re looking for solutions.” But photo contests drop regularly — keeping the feed packed so the real questions drown in the noise. 
