Chatgpt mistakes GPT-5 for GPT-5 Thinking Mini

[half anchor]

Reported by @inner marten

Bug Report: Chatgpt mistakes GPT-5 for GPT-5 Thinking Mini

`Steps to Reproduce`

I sent one message after the actual result, first one was big then tested with a smaller one, but same results, look below to see the results.

`Expected Result`

Expecting GPT-5 to generate it since i had my free 10 prompts.

`Actual Result`

but instead i got a response that said "Thinking longer for a better response", it used GPT-5 Thinking mini without my permission, i thought it was because of a bigger request, so i made a smaller one but i had the same result. and after trying too many times chatgpt said i reached my free limit of GPT-5 even though it was GPT-5 Thinking mini that was responding!

`Environment`

Both app (version 1.2025.245) and Google browser, Ipad air (ios) and OnePlus (android)

Sep 12, 2025, 3:25 AM

Additional Information

Please provide relevant details to help resolve the issue, such as:

• ChatGPT Shared Link (if applicable).
• Screenshots or videos demonstrating the problem.

-# ➜ Need to contact support? Visit the OpenAI Help Center.

[inner marten]

I kept editing to see if it would change if i mentioned it but it seems like it's on instinct

I checked the browser

And Then

[inner marten]

@reef holly @vast bay can i please get help? I can't do anything in my chats

[reef holly]

Hi!

[inner marten]

Hello

[reef holly]

Can you please explain what you mean under *Can't do anything"?

[inner marten]

Take ur time in reading everything

[inner marten]

Can't get GPT-5 answers

Look above

What i reported

[inner marten]

Ts

And the images

[reef holly]

I read everything, but I'm still not sure on what you need help with

[inner marten]

Because i want GPT-5 to answer and not gpt-5 thinking mini..

[reef holly]

If you use your GPT-5 free limit, your model will be switched to GPT-5-mini

[inner marten]

No no no

You don't understand

[reef holly]

This is a model switch, this cannot be circumvented with prompts/system prompts

[inner marten]

I HAVEN'T reached my limit yey

Yet

But when i try to send a request

It answers with gpt 5 THINKING MINI

Not with GPT 5 even though i didn't reach my limit yet

And not even with gpt 5 mini

It thinks longer and uses gpt 5 thinking mini even though i didn't tell it to

[reef holly]

Okay, sorry for misunderstanding

[inner marten]

It's all good

[reef holly]

I believe, in free tier, your messages are forced to use the Auto mode which decides how long to think

And so route your request to mini thinking

https://help.openai.com/en/articles/9275245-chatgpt-free-tier-faq

[inner marten]

But i never had that before

[reef holly]

https://help.openai.com/en/articles/11909943-gpt-5-in-chatgpt

[inner marten]

It happened litteraly just today

And when i asked my other friends they said they didn't have any similar issues

It's just me

[reef holly]

It might aswell be a temporary bug

[inner marten]

Temporary bug from YESTERDAY

[reef holly]

If the issue persist I suggest contacting to help team at https://help.openai.com/

Sadly I have little experience with how to free plan operates

[inner marten]

I found an alternative

[inner marten]

That's fine

[inner marten]

Basically

When it does it

I click stop

Then regenerate

Then it switches to "Thinking"

Then i click skip

Then it actually generates a gpt-5 response

[reef holly]

Keep in mind that the free tier is optimized so OpenAI can scale it to lots and lots of concurrent users

[inner marten]

Yeah ik ik

[bleak basalt]

same

[inner marten]

Is there any staff here that has experience with free tier

@vast bay

[vast bay]

We all were free tier, before March 2023! There wasn't a paid tier of any sort before like the 23rd of that month.

I went plus the day they opened it, I could afford the $20 and wanted ChatGPT to be there into the future, and was scared about the money I feared OpenAI was losing to provide the free service. I wanted to help, just couldn't help much!

So, free has changed a lot in the 2.5 years since I last had a free account - but.

I can use an incognito window and explore what a not-even-logged-in user would see.

[inner marten]

That's cool but do you have any idea on the problem I'm facing

Also I'm logged in

On app and on chrome browser

Everytime i send a request in some of my chats it says "thinking longer for a better answer" and uses GPT 5 Thinking Mini instead of GPT 5 even though i didnt reach my daily free limit yet and didn't use the Thinking mode

Although when I started a new chat the problem faded

So its just in already existent chats

[vast bay]

I don't know about it. You've done the right thing to make a bug reports, to help inform OpenAI; you can also reach out to OpenAI tech support through https://help.openai.com/en/articles/6614161-how-can-i-contact-support and ask them to check your account and see if it needs something adjusted to get you back to normal (I've asked them to help with weird stuff that affected me occasionally).

[vast bay]

Interesting, sounds like a sort-of workaround, and useful to explain clearly in a bug report, because anyone trying to recreate the bug might need to know that.

Also, so, you make a new chat now. If you leave it and come back, is it affected by this 'forced to think' issue, or is it only chats from before some older date that are affected? Stuff like that affects the ability of the techs to find the problem - and if they can't find it, it can never be fixed.

[inner marten]

Only older chats before yesterday

Yesterday at like 11pm est

[vast bay]

So, if the problem was an accident and accidentally fixed (and things like that happen) - then it may 'die out' but take those chats with it.

If you care deeply about those chats being affected by this, you can help explain the bug clearly, define its shape, and with luck the techs may find a fix.

[inner marten]

Yeah bro i NEED those chats

[inner marten]

When i do "Thinking" mode then click skip it generates an actual gpt 5 response

However everytime it says thinking longer for a better answer it actually takes away 1 gpt 5 prompt

[vast bay]

Which means, you care; if you can edit your OP, be sure to include all the details you can for reproducibility, the clues like when it does and doesn't affect chats before and after, may help them figure out what is wrong.

[inner marten]

Chatgpt treats it as a gpt 5 answer

[inner marten]

Do they even look at bug reports

If they can they can just read this entire Convo with me and you

[vast bay]

As near as I can tell, they do at least sometimes. I think it's far less likely they read the entire chat - they're busy people with loads to do. You're the person with a fix you hope happens; I would if I were in your situation, try to make it as clear, direct, short, and easy for the helper as possible. They have something like 700 million weekly users for ChatGPT alone. They probably do not have time to read all chats, or comment 'saw this' or whatever. So, its on you. You say you want a fix; that's how I'd ask for one, by updating the OP to be as easy to follow as possible.

[inner marten]

I cant edit it

Ill just do a new one and put everything in it?

[vast bay]

Understood. Another possible answer, if you choose to provide a clearer bug report, would be perhaps make a new one with the fuller details, yeah.