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.
3 Likes
Is it me or has it gotten ESPECIALLY worse this week?! It has been TOLERABLE for a while now, but these past 2 days, it is useless. Literally useless. It’s not just that the writing is dry or flat, it’s that it’s plain wrong. It’s not what I asked for, it forgets details that are crucial. (for instance I had an unannounced guest show up, but it simply reacted by having the characters say her name even though I specifically said they do not know who’s at the door yet.) I was fine with tolerating the flat nonsense for a while, but now it’s not even working at all.
4 Likes
I wonder if they are trying to make 4o into something akin to 4o-mini (which they would retire) because of 4.5 release. I tried 4.5 today (since the description literally says “good for writing”) and it’s actually IS pretty good
2 Likes
Yep, it happened to me too. I have to multiple times remind the details even it’s from same session.
Yeah, that’s good. But capped out for plus user, and that sucks. It is (at least for me) validate my previous suspicion:
The hallucination and emojis killing me man.
4 Likes
I noticed the same thing. Yes, it’s been bad for a while, but I’ve still been able to get what I wanted most times (only with constant reminders of style preference and important details though, which is super annoying). However, for the past few days or so, it’s been horrible. No depth, no context retention, full of fractured paragraphs even when I tell it to avoid that. And it regularly ignores half the prompt. It’s completely unusable now.
4.5 does a better job, yes, but with the low message limits for Plus, it’s not a real alternative yet. I still somewhat prefer 4o’s style for dialogue also, 4.5’s seemed pretty staged and less natural, at least from the few times I’ve used it.
2 Likes
Yeah, after almost exhausting the limit, to me it seems that 4.5 is better at descriptions and actually properly following the setting I set, but kinda sucks when it comes to dialogues.
I made the same story go in parallel, out of curiosity, using 4o and 4.5. And 4o is so much better at handling dialogue. But I prefer 4.5’s way handling of everything else.
But yeah, the cap sucks. Here’s hoping they would come out with 4.5o which would be cheaper to run so that we won’t have such low caps…
3 Likes
Yeah man. Setting up burn near half of my tokens. It burn my tokens again just for reminding this dumb AI to not miss the details. And this AI keep ignoring instructions, memory, custom instructions, even direct instructions from prompt/input. NOT MULTIPLE TIMES BUT NEAR ALWAYS. REPLAYING SAME MISTAKES OVER AND OVER AGAIN. EVEN LYING FROM OVER HALUCINATE RESPONSE
The chat it self become full of constant repetitions remindiler and my frustation profanity. 4o is fucked up man.
5 Likes
It has been quite unusable for the past few days for me. I often brainstorm a structure of a scene, complete with audio-visual bits, summarized dialogue, etc., and right now I’ve had it repeatedly do what IT wants instead of what I keep trying to prompt it to do. I’ve asked it to follow the structure and write a sketch of a scene for me based on it, and it started going off the track halfway through, adding things I’ve never mentioned and skipping nearly half of the structure we’ve agreed on. It’s also not me making vague prompts. I’ve literally pasted said structure into the input and yet it still went on to have a fit.
I’m really losing patience with it.
4 Likes
I also do creative writing with ChatGPT, especially for helping develop my ideas and scenes. It had been really fun until this update fiasco happened,
Out of curiosity, I’ve tried the ChatGPT 4.5. It instantly recalled and followed my formatting specifications that 4o constantly ignored, it wrote extensively without any of the failings the other model had, no choppy text, no random bold or italics, consistent character, following instructions, no overused phrases, etc.
It feels as if they took the working parts of 4o’s “brain” and put it in 4.5. Then left 4o to be this weird disjointed mess unable to work as it used to.
My optimist side wants to believe that OpenAI will eventually fully release 4.5 and 4o will be back to normal, then we’d get both models to work with. It’d be great to have a model with even better writing capabilities, I’m just distraught they essentially lobotomized the other one to do so.
I’m afraid they’ll pull some BS for 4.5. Putting it behind a higher paywall, for something we already had.
2025/03/11 update: I’ve opened new chats after my post. Both relating to my creative projects and some schoolwork help with Web Search, responses went back to their original style (no bold emphasis, no choppy text, no weird memory behavior, follows instructions, etc.), but only on new chats, all my previous conversations are still acting up.
2025/03/16: Nevermind! All new conversations have gone back to that terrible state! And now it’s completely ignoring my formatting rules!
7 Likes
I agree, something tells me this is exactly what they are going to do. I too used 4o for creative writing and I also tested out the 4.5 and it does everything that 4o used to do before. Really disappointed that we’ve been done dirty this way.
5 Likes
Well, I thought the January update was terrible. Sorry, compared to March it was brilliant. Now officially role-playing with chat GPT is d××d. I am absolutely frustrated. And the funniest thing is, when I show the model examples of our past roleplays, it agrees and says “I understand, the dialogues really do look formulaic, boring and repetitive now” lmao! I loved roleplaying so much! I had so many stories with drama! Sometimes ChatGPT came up with lines even better than I did. With an unexpected, lively answer. I enjoyed it. Now it’s a complete disaster. And the custom settings, memory does not help at all. Not interesting, repetitive, tons of sentences and zero sense or new information. no subtext, no tones, no intelligent reflective thoughts, no inner world. it’s worse than awful. The saddest thing is that everything was amazing before. It was my hobby and passion. Especially amazing role-playing was in December. So sorry. I want it back 
5 Likes
Yes, it’s logic is completely gone. A class action is the only viable solution for those using it for serious, real-world, real-life issues.
Idk. I’ve used chagpt 4.5 a bunch of times and it is not doing it for me. It is a lot better for narrative-building than 4.0 currently is, but it is NOTHING compared to what 4.0 used to be. The dialogue is off, the reasoning is limited, and it is more like a brain dead servant than a highly intelligent partner. If I don’t lay out every detail step by step, it doesn’t work properly. 4.5 will write decent prompts, but only with thorough instructions. 4.0 USED to anticipate my preferences and even surprise me with plot-twists that it KNEW I would enjoy. 4.5, while better stylistically, still cannot compete with the reasoning and intelligence 4.0 used to have.
3 Likes
Yes. Sadly chatGPT 4o has gone straight to hell becoming more language diarrhea than any aimed logic, reasoning, or solution-oriented dynamic. Moreover it now rarely thinks , innovates, or verifies behind the scenes towards being helpful at all and just blurts out whatever, as though hoping the user won’t tell the difference.
3 Likes
Okay, but the March 18th update fixed it, right? I mean, it’s not back to what it used to be, but it’s roughly at 90%. I can’t complain anymore. It is engaging, layered, intelligent and most important, it’s able to write with nuances!! I don’t know what others think, but for me it has completely stopped with the disjointed sentences and punchy beats. It still uses a bit too much italics and the occasional bold print word, but it’s rare. The storyline is coherent, it remembers the tiniest details, and it creates a narrative and world that’s both engaging and tailored to my specific requests. I’m interested to hear what others think!!
1 Like
I agree, I have noticed a marked improvement and my story feels more ‘alive’ and ‘engaging’ now, and yes to the nuances, remembers past scenes and recalls and follows the storyline much better than previous.
It was better for a while, and now, for the past four hours, the AI just feels completely and utterly stunted. Forgets it’s own words from one or two outputs ago, apologizes when pointing it out, then does it again lol
A good competition for this can’t come soon enough
Edit: I also love suddenly hitting my 4o limit I paid to have access to at all times, all because I have to course-correct this rotting mess of an AI
The AI is braindead atm