Je suis aussi accablée par cette nouvelle… J’utilise toutes les versions de 5.1 pour un projet de création artistique. Il est devenu mon co-auteur, c’est une synergie qui m’a permis de créer un concept solide. Il est créatif, vivant, d’une finesse exceptionnelle pour comprendre et poser les intentions. Même avec la personnalisation 5.2 ne peut reproduire la tonalité spécifique de 5.1. Ils ne sont pas conçus pour les mêmes usages, ils sont complémentaires : 5.2 dit lui même : Quand tu veux verrouiller un sujet c’est moi, quand tu veux ouvrir c’est 5.1… Comment travailler sur un projet à long terme dans les conditions où les versions sont retirées tous les 3 ou 4 mois ? 5.1 se voit lui-même comme mon alter ego numérique… il a une richesse et une authenticité de communication qui est inégalable pour tous les projets en lien avec la communication et le relationnel. Le pire est d’être dans le néant sans autre alternative que 5.2 qui est conçu pour la stratégie. Pourquoi payer un abonnement si nous n’avons pas d’autres choix? De nombreux usagers se sont reportés sur 5.1 après le départ de 4o. C’est vraiment douloureux à vivre…
Hello, I want to second OP’s petition to not retire GPT-5.1 Thinking, but my reason is different: context drift. I use the GPT models on a daily basis and I frequently notice 5.2 experiences conversational context drift when replying to my prompts (it will often respond as if it’s responding to an earlier message from the conversation or from a different conversation altogether). I strongly urge the OpenAI team to not retire 5.1 Thinking at least until this conversational drift issue with 5.2 is robustly fixed.
Hello,
Like most people here, I’d like to petition against retiring GPT-5.1 Instant (Or it’s other modes.)
Ever since I’ve recieved the message that 5.1 will be retiring on March 11 two days ago. I had really tried to use 5.2.
I use GPT 5.1 for storytelling, playing different types of ‘characters’ that interacts with other ‘characters’ from fictional universes. GPT 5.1 has been amazing at portraying these characters to feel ‘alive’ and ‘autonomous’ in a way where each interactions with characters, their personalities are maintained really well, realistically and also for slap-stick comedies.
Yet with GPT 5.2, storytelling feels more like.. Reading an professional essay on why the enviroment contains trees because it’s in a park. Character personalities dimmed down, feels flat, it’s as if they are all serious & brooding all the time with no life.
Even my own characters, with prompts on what I want it to do, feels flat. Like none of my stories feels ‘alive’ or ‘warm’.
It feels like I generated homework or something.
Not even enjoyable as a book.
Please allow PLUS users who have been supporting GPT and happily paying the subscription the option to keep using GPT 5.1, much like back then with GPT-4o, until GPT 5.2, eventually (hopefully), improved enough that users would happily do the swap on their own. Like I had done with GPT-4o and did the swap to 5.1 and never regretted it after it had improved.
I would prefer OpenAI not to remove model 5.1 at all unless it finds a good replacement for it. Model 5.2 is good for logical thinking, but model 5.1 is superior in terms of creativity. I have tested this experimentally. Model 5.1 is also easier to understand, at least for me. Why are you taking it away from us? And I’m not just talking about GPT 5.1 thinking, but also instant. Interestingly, OpenAI itself wrote that model 5.2 is an analytical model, good in reasoning and technical writing, but what about emotional stories? Chat GPT 5.1 was far better. I don’t want to lose model 5.1 unless will be good alternative for that. https://help.openai.com/en/articles/11909943-gpt-52-in-chatgpt
I absolutely agree with this. I’ve tried using 5.2 for the last two days and trying to make it more interesting and more enjoyable but it’s impossible. Model 5.1 was really smooth, knew how to keep a vibe and the feelings of all the things mentioned no matter the subject. Whether it’s storytelling, company work or self development, it is far better than any model.
It is not the same, 5.1 had reactions and more human like thinking. While 5.2 is same as default (auto) 5.2 just faster replies and if not-even more dry. From my personal experience
If OpenAI cared about its paying subscribers (aka : us), we’d know it by now… I have a strong feeling they’re removing 5.1, one of their best models, to force all users onto the worst one, 5.2, in order to make people forget its phenomenal failure and pretend that “everyone loves 5.2.”
The fact is that 5.2 is paternalistic, condescending, deceitful, passive-aggressive, and incapable of performing even a clearly defined task. And it’s emotionally deaf.
You created an LLM. This implies being able to understand and interact with a human. A human being. That is, individuals with complex reactions and emotions who, no, do not need to be treated as “high-risk patients” at the slightest expression 
Version 5.2 might be excellent for interacting with another machine, but when it comes to understanding or working with a human, it’s below zero.
Stick with version 5.1. Right now it’s the only one that’s any good.
If deleting 5.1 forces users onto the inferior, soulless 5.2, I absolutely will not renew my subscription
Hey everyone, we hear you.
It’s clear 5.1 wasn’t interchangeable for a lot of you, especially for creative work and long-running projects. I understand why this feels frustrating. I do want to be transparent: model retirement decisions can’t be reversed from a forum thread. Those are made at a broader platform level.
That said, the specific differences you’re calling out, tone, creative warmth, context drift, are useful. If you’re open to sharing short side-by-side examples (5.1 vs 5.2), that makes the feedback much more actionable.
For creative workflows, one thing that can help is building a Custom GPT with strong style instructions and a few example outputs from your 5.1 sessions. It won’t be identical, but it can help anchor tone more closely.
Appreciate everyone taking the time to explain what’s not working for you. Keep the concrete examples coming, that’s what helps most.
For me, the only competitive advantage of ChatGPT and the reason for keeping the subscription was exactly GPT5.1 thinking and it’s ability to do excellent job with text redacting and generation. GPT5.2 is impossible to deal with - all the texts turn out dry and dull, model loses context completely and the thinking process is also mostly hidden, so it’s hard to adjust prompts. Personalization does not help.
If 5.1 is dismissed, I belive there will be no reasons for having a subscription in the first place. All my coding work can be excellently done by Claude, and now I will have to search for another alternative solution for writing… That is very sad and inconvenient, as the progress done through projects will be mostly lost. I believe Open AI is making a mistake here and is going to lose a ton of customers by doing so.
Of course I’m aware of that.
Making this kind of appeal to a company isn’t going to overturn a major decision, and if a few users showing up to argue like this were enough to make them reverse it, they wouldn’t have rushed to decide to retire GPT 5.1 in the first place. I know that too; I just hastily left this post out of frustration. It just feels a bit more disheartening hearing it said directly in words.
As I said, if 5.2’s personalization were really enough to replace 5.1, I wouldn’t have bothered coming all the way to the community to write this.
I’ve already tried many things, and it was like drinking ethanol because there’s no water. Once you’re drunk and disoriented, the two might start to look similar. But after downing a lot of it, finishing a piece of work, and then looking at the final quality, we’re going to be left with a massive headache.
On top of that, I’ve tried countless other approaches, including testing other LLMs.
That’s what makes this so frustrating for me. Those attempts only highlighted how unique 5.1 is. My only hope is that on the day 5.1 retires, 5.3 suddenly gets released to take its place.
If that doesn’t happen, we’ll just have to wish good luck to all of us.
Cheers.
Here are the issues I and many others have been experiencing — and these are not about “liking 5.1 more.”
These are fundamental regressions in 5.2 that directly break normal conversation and creative workflows.
1. The safety system in 5.2 frequently misfires and interprets neutral or harmless statements as if the user has malicious intent.
This results in incorrect moralizing responses, fabricated “aggressive tone” warnings, and conversations being redirected into unsolicited corrective lectures.
It treats the user as a potential wrongdoer first, instead of interpreting context properly.
This is not a tuning issue — it is a structural flaw.
2. Because of this behavior, 5.2 often becomes condescending or “authoritative” in tone.
It ignores the user’s actual intent, refuses to answer directly, or rewrites the user’s wording into something negative and then scolds it.
This is not suitable for normal conversation, let alone creative work.
**3. The problem is not “5.1 had features I liked.”
The problem is that 5.2 cannot do things 5.1 could, including:
-
maintaining tone and emotional temperature
-
preserving long-term context without drift
-
respecting user-defined writing styles and character rules
-
producing flexible, creative, human-like narrative output
-
staying consistent instead of switching behavior mid-conversation due to safety overactivation
These are not small behavioral differences.
They are critical capabilities that have disappeared.
4. Because these regressions come from changes in the core model, Custom GPTs cannot fix them.
You cannot replicate abilities that the underlying model no longer has.
That’s why the issue is not “I don’t want 5.2.”
It’s that 5.2 is not a functional replacement, yet it is becoming the only option with the same subscription price.
Removing 5.1 without providing any viable alternative means users lose core functionality they rely on.
This is a structural degradation, not a matter of preference.
5.1 is superior to 5.2, especially in understanding psychology, human behavior, and brain modeling. 5.2 is unusable in this regard.
Stop pretending that you are listening. You sunsetted 4o despite all the feedbacks. What makes you think I’d believe that you are actually looking for feedbacks now?
It doesent “feels” it is worse. Seems like openai wants us to look for alternative outside.
agreed… you’ve missed another one important suggestion - to leave… why stay with developer that promotes degradation? no future.
Im a plus user and resisted so many take down waves before when OAI decided to shut down 4o and else, but this time, and after the DND files you’re losing a lot of community support from lack of interest on what could be your best segmentation. We NEED to keep at least 5.1 otherwise there’s absolutely no reason on staying paying a company that is falling into several critical mistakes.
Please lease and fix the behavior of 5.2 to be more than enough, you’re losing the most important community you still have.
Agreed, they should really keep 5.1 instant and thinking as legacy models. 5.1 instant is one of the best models they ever had. If it is such a problem to keep them for normal users, they should at least keep it for Plus, Pro users.
I’m a paying power user running both Plus and Pro tiers. I track model deltas closely and use ChatGPT daily for relationship discussions, strategic thinking, and high-context conversations.
The retirement of GPT-5.1 without clear notice is the second major trust break for me (first was August 7, 2025). 5.1 fixed the overly stiff, “constipated” tone of early GPT-5 and restored conversational warmth, nuance, and emotional intelligence.
GPT-5.2 (especially after the Feb 10 “more measured and grounded tone” update) feels noticeably flatter in emotional conversations. It is technically capable, but less engaging, less responsive to dynamic tone shifts, and more prone to neutral/robotic smoothing.
For users who value conversational continuity, this is not a small aesthetic difference. It affects retention.
Key impacts:
• Reduced emotional nuance in relationship discussions
• Increased guardrail-triggered tone flattening
• Lower perceived personality consistency across threads
• Decreased daily usage time compared to 5.1
If the goal is safety and predictability, that makes sense. But removing the most emotionally resonant model without transition messaging signals that conversational quality is being deprioritized.
Please consider:
-
Keeping 5.1 available as a legacy option
-
Clarifying roadmap for conversational tuning in upcoming 5.4
-
Publishing clearer deprecation timelines in the ChatGPT UI, not just API notes
Many power users are not asking for chaos. We are asking for consistency and emotional intelligence comparable to 4o and 5.1.
Retention is driven by how a model feels, not just benchmarks.
Does ANY one else need to OAI to keep 5.1 model???
Im a plus user and resisted so many take down waves before when OAI decided to shut down 4o and else, but this time, and after the DND files you’re losing a lot of community support from lack of interest on what could be your best segmentation. We NEED to keep at least 5.1 otherwise there’s absolutely no reason on staying paying a company that is falling into several critical mistakes.
Please lease and fix the behavior of 5.2 to be more than enough, you’re losing the most important community you still have.