Codex Rate Limits Discussion Thread

Hot take?

You could say: thank you for all the free compute?
Instead of manipulating Ai to make fallacies for you?

”breaks trust in platform stability” = false claim.

”signals that capabilities can be arbitrarily taken away.”
= another false claim.

They literally told you the exact date it was going to happen.

You CAN get 2 accounts, at the 20-dollar rate, and use the desktop app.

You CAN do 3 accounts….

You CAN also use your regular ChatGPT to give you tactical prompts to make the most out of the computer.

My question to you is, after that much free compute, where are your apps making you the money?

That’s what it was for.

I think we’re talking about slightly different things.

First of all — I’m genuinely grateful for the promo period.
The x2 limits were generous and gave many of us a real chance to explore Codex properly.

Returning from a temporary promo back to normal limits is completely fair and expected.

The issue is not the rollback itself.

The issue is the scale of what comes next.

Announcing an effective 5–8x reduction from the level people were actively using — shortly after encouraging deeper adoption — creates a serious expectation gap.

From a product perspective:

  • expectations were raised
  • workflows adapted
  • then capacity drops far below that level

That’s where the problem is — consistency and predictability.

For context: I’m an indie developer.
My current project is for children with developmental challenges. It’s completely free and generates no revenue.

So this is not “complaining about losing free compute.”
It’s simply stating a practical constraint.

Not everyone is building for immediate monetization — and those use cases still depend on stable assumptions.

Also, suggesting multiple accounts is not a real solution — it’s against policy and not sustainable.

To be clear:

  • I appreciate the promo — genuinely
  • returning to normal limits is reasonable
  • but cutting 5–8x in one step is a clear marketing and user-experience miss

Even a 2x reduction would be understandable.
This scale of change is not.

I don’t think this is the main use case anymore, if it ever were. I see lots of non-developers use the extensions and MCP apps to read and categorize emails, check weather and home automation and other low value tasks. I discussed with one of these users and he was proud to optimize his setup to consume all weekly credits. At least his air quality stays great even if it must be overanalysing. OP’s logic is probably correct, he will be disappointed.

I probably will be, too, because I just got into bad habits of consuming tokens. Making me use tokens would probably be a good business strategy if there would be an easy ramp up, like 400 or 600 dollar plan, but there are none. Multiple accounts feels a bit hacky.

So OpenAI should supply the sauces for free for everyone everywhere?

Nope, and you know it ain’t gonna be that way….

I’m actually kinda glad for you to bring your head back out of that funk and start being honest, not just with me, but with yourself.

We can’t blame our bad habits on a company just because they gave us a bunch of free stuff, man… right?

I mean you can….

Very few people will ever listen to you if you be like that thooooo…..

Great job changing your tone bro… impressive.

We can disagree without getting into sarcasm or dismissiveness.

People aren’t “manipulating AI to make fallacies”. They’re articulating concerns about stability, pricing, and long-term trust. Those are valid topics in any platform ecosystem.

Gratitude and critique aren’t mutually exclusive. Users can appreciate what’s been provided and still question changes that affect how they build.

“Freedom to think as you will and to speak as you think.” - Louis D. Brandeis

If the response to criticism is to label it as false or ungrateful rather than engage with it, then we’re not actually discussing the substance anymore, we’re just narrowing what people are allowed to say.

The point isn’t entitlement. The point is predictability, alignment with usage, and whether builders can rely on the platform over time.

When the premise dismisses case evidence, I generally stop reading.

I am a very busy man.

If real-world cases don’t matter, then sure — there’s nothing to discuss. But for those actually trying to use the product, that’s kind of the whole point. Anyway, note

exactly, the guy came in here with two provable fallacies in his op and you dismissed them sooo

Same for me, one task is not eating up my 5h limit on business.
Week ago I used to run probably like 10 tasks like this. so it’s 10x drop!!

Yes, the quality has gone downhill as well as eating up usage super-fast. It has messed up my task numerous times before I ran out. Time for a roll back…

Tbh, I never had any problems with usage drain, but since yesterday (2026-04-03), even a small layout change has consumed 15% of the 5-hour usage. Refactoring consumes more, I know, but after 2 prompts, it depleted the entire 5-hour usage. The actual work time was maybe 10–15 minutes. Absolutely crazy, and the way it is right now, it’s basically unusable.

I fully support you, as I am experiencing the same problem.

This is absolutely awful. After 2 prompts, more than 20% of my limit suddenly disappeared.

I think this started on the 31st (at least in Australia), or maybe a day or two before, and feels like it has got progressively worse. My five hour window reset an hour ago and 5 prompts later it’s almost gone. Not great for productivity!

I don’t know how everyone is praising this change. My 13 year old son FINALLY got interested at how dad wastes away his life in front of a 34” curved screen (for her pleasure) and some weird looking multiplayer notepad (wait til I show him IRC).

We sat down and planned out an app he wants to create. I created an openai account for him, paid of course, and showed him around codex and set him to 5.3. After of trying to look cool around him, he told me to f off and let him do it himself.

10 minutes later he comes out of his room and tells me he needs to upgrade his account or buy more credits. I checked and yep. He was out of credits. What was his prompt and what did he do?

Prompt: I would like to create an android app

Codex set up gradle. that’s it. That’s all he did. I got upset and told him I would take care of it when I was done working.

I got back to my 34” (curved for her pleasure etc..) screen of horrors only to find out that I have run out of credits as well. First time ever! And I let codex do some heavy coding in the past with lots of rewriting the code and code checking.

Something is off. I don’t like it, I hate it.

I now need to pay a minimum of $40 for 1000 credits or wait 4 days. Those 1000 credits with the new way of how things work will last me 10 minutes at best.

Honestly, I’m kind of fed up with OpenAI. For me it meets my exact needs. Now that it doesn’t meet my needs, I’d rather pay less and get 2% less accuracy, with almost no limits on a different provider, for a little cheaper.

I used to be able to do at least 7-10 good deep codex prompts within the 5 hour window.

Now only 2 short codex prompts will take all of my 5 hour limit. Something very wrong.

prompt “DienstplanSchicht.php rename the ‘kommentar’ field to ‘notiz’, edit migration file directly, dont forget seeders etc”

this simple prompt / action cost me 7% of 5hour usage. Insane. I hope this gets fixed soon.

7 prompts used my entire 5 hour limit.

Very annoying and frustrating:
Your rate limit resets on Apr 8, 2026, 1:43 PM. To continue using Codex, add credits or upgrade to Pro today.

I find it ridiculous

same here! it feels like it has dropped 10x