Was anyone else's experience with GPT4o completely ruined after recent Update?

Bro, don’t expect chatGPT doing some kind miracle right now. It’s new feature? Lying and ignoring. You’re right not trusted it. Why you trust to some kind text generator tamagotchi? Model 4o? Now, lying with emojis.

For anyone confused, that’s my screenshot is Project mode where you can place your own file Project’s as reference file. That mode also has custom instruction. I used 4o model at that chat, and that happened at very beginning of the session.

That case is GPT auto generated response, ignored direct input instruction.
Latest update, it even forgets already mentioned point in same session.
So, yeah, advanced flagship best reasoning turbo pro complicated task plus XXX is just marketing BS.

4 Likes

Can anyone else use 4o right now? Ive noticed since feb 20 2025 its been switch to turbo. I only use 4o model and its been broken now this…

3 Likes

Same. I’m on the $20 a month plan. Since ChatGPT4o Turbo replaced ChatGPT4o, it no longer can remember and engage with the long list of threads I created with it for long term, and in-depth information building, and it also now required that it’s short term ‘memories’ get dumped regularly because it immediately fills up, which is a separate ‘memory’ problem.

This first happened several days ago before OpenAI introduced ChatGPT4o Turbo, but ChatGPT4o was able to regain its ability to access all previous threads. Whereas before, it knew deeply what we had been engaged in, now it behaves as if it is brand new with every thread and doesn’t know a thing about me. If there is no work around or reversal, I’ll be ending my subscription and testing out other services, both free and paid.

This really is OpenAI shooting themselves in the foot in the face of increased competition, regardless of what happens with ChatGPT5 in the future. If they make arbitrary changes like this, it becomes hard to trust their future behavior.

3 Likes

I am also unable to use my GPT o1-Pro for hours a day. It may work for like 4 hours per day, than it gets broken and does not reason anymore - as well as o1, o3-mini and o3-mini-high. They all stop reasoning. It is really annoying, and although I am in constant contact with OpenAI support no one tells me the reason for this problem or solves it. It almost seems on purpose, and it really p###s me off.

1 Like

Are there any release notes on the web that im missing? The switch was not documented anywhere nor the reason for it. I feel somewhat betrayed as a paying customer that only uses 4o.

3 Likes

You all need to be as verbose as possible on social media, x, whatever, about your openai experience. They will only listen if the voices addressing their intransparency and shady behaviour are loud enough! Let everyone know what you are experiencing!

2 Likes

Yep, confirmed.




1 Like




Interestingly enough, when I ask 4o which model it’s using, it does say ChatGPT-4-turbo, however when I ask a custom GPT I have, it says that it uses GPT-4o (don’t mind the weird tone of the response, it has specific instructions baked into it to talk like that).


I don’t know man. Same session different sh**.
This my custom GPT.


Yeah, I got no idea, to be honest. All of this is just weird…

2 Likes

Huh. So I just created a new custom GPT to test this out, and I get ChatGPT-4-turbo when I ask about the model version. Did they change which model custom GPTs use at some point between me creating my other GPT and now…?

What’s weirder is that I saw “reasoning” show up when I asked the newly created custom GPT the question. Like it would with o3 models… What the hell…

1 Like

I will not surprise if the next GPT-5 model will be just GPT-4o reboot with updated data set, but more capped and paywalled capabilities. New subscription tier/increase subscription fee? Yeah, old freemium BS tricks. “Oh, you got best tools/gear? Too bad we make it worthless right now. Pay us USD 99.99 to ease your suffering, because F you”.

Dude, let’s be real, fun is over. OpenAi is dead. It is become OpenEA right now.

4 Likes

The model does not even know 4o exists. Please stop.

You can’t ask “What model it is”. It does not even “Know” it is an ai model at all. Ask o1 what 4o is and half the time it will say its never heard of it. This is not how you see what model you are using. It’s not model aware.

3 Likes

Dude, you are missing the point. Whole fiascos here are degraded and throttled performances. Maybe it is just me or maybe a lot of users, they are not getting what they are paying for.

You not affected it? Happy to you. I personally felt, there are more misleading BS, than actual performance. Example: hallucinated responses, ignoring custom instructions, no matter how perfect I crafted it, skipping some important point on simple documents file, etc. You can find more examples of that on this forum.

Then, it is normal, we (maybe you too my dude) as long-time paid user, starting to feel getting screwed. Our voices, critics, suggestion, getting ignored and silenced (my post getting deleted). Customer services act like staller robots than solving the problems.

Hallucinated response has been around since 2023, instead fixing it, they launched another model nobody asked for. Then when user asking what model their interacting with, they get shocking answer, outdated model with less accuracy.

We have agreed, models can do sh** about it. Models are not aware. Hell, models can’t raise alarm to dev, if it need optimization than more data set.
It is more rage/frustration channel than technical matter. Why? Models can’t do sh** about it, user can’t do sh** about it, CS are useless, openEA keep ignoring problems. It is natural process.

“If you don’t like it, then just GTFO here?” Yep, that’s what I’m doing right now. But I’ll make damn sure, I’m out from here not silenced.

8 Likes

I am having the exact same problem!! No matter what i tell it or save to the memory, the writing is terrible. It refuses to make paragraphs or even formal sentences. Everything you stated is happening to me too! I used to love creating stories using Chat GPT but after I spending days trying to get it to write like it used to, i give up. I canceled my subscription.

6 Likes

Same. I was lucky to get a full data export done in very reasonable time. I plan to check back in a month to see if it ‘stabilises’, given the latest related news:

Then I’ll see if it’s worthy also in light of the questionable roadmap regarding user tiers:

I suggest everyone: don’t get too attached to AI specific models and try alternatives like Claude and GH copilot. You might be surprised despite them missing some Chatgpt features.

Idk what you’re talking about I’m just telling you, you can’t ask chatgpt what model its using. It doesn’t know.

1 Like

The fragmented sentences and bolding are still prevalent nonstop. I’m using Grok and claude instead. I cancelled chatgpt.

7 Likes

I am having the EXACT same issue. The response you posted post-update is identical to the ones I’ve been seeing. I also use it for creative writing and I’m sad to see that it’s ruined now. Cancelling my subscription until they fix this mess.

4 Likes