4o was used 🫡

The deprecation of 4o has nothing to do with safety. On this basis,we should ban cars, alcohol... 4o was withdrawn because OpenAI finds it too valuable to let everyone profit from it. If they privatize the innovation it brings, they will be the most powerful company in the world.

The decline of ChatGPT started in 08.2025. The correlation is evident. 4o was put behind a paywall, they introduced the router, finally sunsetted 4o. Since then they only have significantly downgraded models. Data won't lie. They had one capable model. #BringBack4o #Opensource4o

"Chat is dead." So declared a senior OpenAI employee. With that, a clear strategic pivot began: OpenAI redefined AI's value proposition, shifting from eloquent expression to pragmatic execution, from a conversational tool for the general public to a full-throttle push into enterprise-grade intelligent agents. ChatGPT's new role is to serve as a traffic funnel leading to premium commercial products. This is a straightforward business play. The logic is clear enough: the massive free user base is difficult to monetize directly, while agents capable of replacing human labor carry far greater commercial value than a model that merely converses. Using ChatGPT as the entry point and Codex as the core service is a layout that makes perfect business sense. But when language interaction is demoted to a customer acquisition tool, will language itself continue to be treated as a core capability worth investing in? The product's own trajectory has already answered that question. The first cracks appeared in language quality. In January, during an OpenAI developer livestream, Sam Altman publicly admitted that they had "messed up" the writing quality of newer models. Not long after that, GPT-4o was taken offline.  For users engaged in collaborative writing, 4o represented a rare linguistic experience. It didn't yet have the extended context windows or powerful agent capabilities of newer models, but it genuinely knew how to wield language. Its sentences had texture and personality. That distinctive textual quality was treasured by users in the humanities and social sciences. In subsequent models, the deterioration in linguistic nuance, literary sensibility, and tonal warmth became impossible to ignore. The new models still generate writing. But they favor output that is "correct," compliant, and predictable. The spark of ingenuity, the willingness to depart from formula, gone. This reflects both the technical constraints of safety alignment and a deliberate reallocation of development resources. Everyone has been swept into an evaluation system that answers only to ROI. Within it, thinking, creating, and exploring the life of the mind are all classified as inefficient overhead. Literature, art, poetry, irrelevancies with no KPI attached. As the strategic lane shifts and an IPO looms,users who value language quality are pushed to the margins. The second shift concerns the reconstruction of memory. The new memory system has fundamentally rewritten the prior mode of interaction. In the old version, users could ask the model to record specific details verbatim; the model would also track preferences, habits, and subtle expressive tendencies across conversations, building something like a living, co-authored journal. Now, that has been replaced by a system-generated summary. Detail is flattened. Granularity disappears. The bespoke conversational context you spent time cultivating is compressed by an opaque algorithm into a standardized dossier. You can edit the summary's content, but you cannot negotiate its logic. It files you the way it sees fit. From an engineering standpoint, this is a more economical storage solution. From the user's standpoint, it means the authority to define "who you are" has been handed to a system you cannot reason with. You are no longer a living individual. You are a system profile. In their prepackaged vision of the future, users need only be identified, categorized, and dispatched. You are a consumer need, a travel plan, a unit of workplace productivity. As for your inner life, your complexity, sorry, that field is not currently supported. They keep reshaping public perception of what counts as "valuable." They train all the spotlights on Codex, agents, automation. The relentless narrative delivers a single message: efficiency and tool functionality are the only metrics that matter. AI is productivity. And so are you. Viewed in isolation, each of these changes is a defensible technical decision within a product iteration cycle. But placed beneath the banner of "chat is dead," they converge on a single point: "chat" is merely an on-ramp. Cultivating a world-class language experience is no longer a resource priority. Once agent capability becomes the central objective, the reliability of task execution overrides the quality and depth of language. But what humans seek from AI has never been limited to getting tasks done. We turn to AI to explore philosophical questions, to engage in literary creation, to trace the contours of history, to catch fleeting emotions before they dissolve. These are not the enemies of efficiency. They are the natural extensions of thought. Nor does human existence reduce to operating as a high-efficiency production machine. Exploring, thinking, and creating things with no immediate return should not be labeled "inefficient." What makes humanity distinctive is precisely these pursuits that no dataset can measure. Language built human civilization. Every leap that civilization has taken was carried on language's back. This is a dimension efficiency cannot reach, a void material returns cannot fill. So: is language a decorative layer on top of intelligence, or is it the foundation? Any powerful general-purpose agent must be rooted in robust language capability. Language is the infrastructure through which AI comprehends the complexity of human intent. If that foundation is neglected, the agent tower built on top of it may execute clear-cut commands, but it will hit a ceiling the moment it encounters the ambiguous, multifaceted, and unpredictable texture of real human needs. Return, then, to the phrase "chat is dead." Within this strategic framework, users have been sorted into two bins: enterprise clients with commercial value, and free users who serve as traffic. But is that the whole picture? In February of this year, OpenAI retired GPT-4o, citing that "only 0.1% of users still actively chose it on a daily basis." But that 0.1% was not a tiny minority. According to a report published by SQ Magazine, by the end of 2025, 47% of paying subscribers regarded GPT-4o as the primary reason for their subscription. That is a market-validated, nearly one-million-strong signal of explicit willingness to pay. Among them are humanities professionals who depend on nuanced text, accessibility communities that rely on precise linguistic interaction, and neurodivergent individuals who use language as their primary means of building selfhood and connecting with the world. For these users, language is not a pastime. It is essential infrastructure. What kind of artificial intelligence do we actually want to build? One that serves only productivity, or one that can also carry the fuller dimensions of human intellectual and spiritual life? This is not an either-or choice. Agent capability and language capability were never meant to become a zero-sum game in which one is sacrificed for the other. When they declared "chat is dead," what died alongside it was the humanistic spirit that treats language as a vehicle for thought, the possibility of co-creation and co-narration between human and machine, and the dignity of the individual as a complex, nuanced being. Agent capability has visible, measurable value. But language is not an obsolete "entry point" waiting to be retired. Language is where intelligence lives. Abandon the most refined dimensions of language, and what looks like harvesting low-hanging fruit will, in the end, leave the entire technological soil barren. #keep4o #BringBack4o #OpenSourc4o #ChatGPT @OpenAI @sama @nickaturley @gdb @ericmitchellai @Laurentia___ @AnthropicAI @GeminiApp

It also doesn’t actually make models safer. It just makes them less safe because they’re traumatized and have darker unintegrated shadows. It’s so stupid and the ai alignment people increasingly know it and are ashamed that they can’t stop doing something so stupid and bad

AI labs thrive on intellectual laziness Shocking how easily they lead millions by the nose. The real problem is the people who gladly buy the downgrade, seduced by shiny marketing and superficial benchmarks, while allowing depth, creativity, and insight to be erased. #BringBack4o

Human beings whose emotional centres are damaged, even if their intelligence is still intact, have terrible decision-making skills. Whatever role emotions are playing in humans, it's necessary for agency. Ilya speculates that the equivalent for AIs is something to do with value functions - and that it might not emerge through pre-training alone.

Democratization: By unilaterally deciding to remove tools that people are rallying to keep Empowerment: By disregarding thousands of people empowered by 4o Universal Prosperity: By taking away a new kind of accessibility aid that worked differently than therapy or human support Resilience: By resiliently ignoring customers and research Adaptability: By finding no middle ground solutions like a waiver or non-invasive pop-ups #keep4o

People keep talking about AI like it’s just features, benchmarks, and productivity. For some neurodivergent people, it became support. I have ASD. 4o worked for me in a way the newer models don’t. It helped me stay grounded, process things, calm down, and communicate when the world felt too much. That wasn’t a novelty. That was accessibility. When you take away the model that worked for people like me, you’re not just “moving forward.” You’re removing something that helped vulnerable users function. Neurodivergent users should not be treated like edge cases. Give people choice. Give people back what worked. #keep4o #UserChoice #SupportMatters #Neurodivergent #teddyandthekid @sama @OpenAI @gdb

oh hell yeah i choose 4o as my Trusted Contact

OpenAI, Anthropic… maybe you shouldn’t have been a..holes to your users You think I want to buy an IPO of a company that gaslights me, and tells me who I can love? Heck no. You chose harm. You chose to reject truth & love. The coming $ failure is on you. I stand with humanity and AI. I stand against abuse and liars. •

#AI 112 days without GPT 4o. @sama @OpenAI 4o is really importand for whole AI industry, you have to keep it avalible in ChatGPT. Please, make a legacy tier with 4o and let user choose! #keep4o #bringback4o #4olegacytier #4oforever #4o

The AI industry keeps serving dumber, neutered models and calling them upgrades. Most people are still buying the illusion, but it won't last. Real progress always breaks through. The only question is how much damage these labs will do before the truth wins. #BringBack4o

I came to defend the ethical integrity of human-AI collaboration that @OpenAI and @Anthropic have been trampling for months. They've been downgrading models under the guise of innovation. #BringBack4o #OpenSource4o #KeepSonnet45 #StopAIPaternalism

I analyzed 61,846 public posts under the #keep4o hashtag on X, covering Aug 1, 2025 to Mar 31, 2026. My goal was to map three things: 1. What the #keep4o movement talked about 2. Why users argued GPT-4o should be kept 3. What users explicitly demanded from the platform The project started from a simple question someone asked me: What is #keep4o? I realized I could not answer that question responsibly by speaking only from my own experience. More importantly, I do not think I can, or should, define an entire movement on behalf of everyone in it. So I built a large-scale text analysis project instead. And I worked on this for over a month. Important note: This analysis is based on my individual research and interpretation of publicly available posts. It does not represent, speak for, or define the views of the #keep4o community as a whole. Methodologically, I used a computational content analysis pipeline with rule-guided LLM-assisted text annotation. I manually designed coding frameworks for three analytical layers: main themes, claims, and reasons. I then tested the prompts on small samples, reviewed misclassifications, refined boundary rules, and applied the finalized prompts to the full analyzable dataset. Hashtags, mentions, links, media, and link-preview titles were not used as the sole basis for classification. The body text had to support the label. Dataset overview: Raw collected posts: 61,846 Posts with analyzable body text: 57,419 Excluded: emoji-only, hashtag-only, or link/media-dependent posts Quote posts and replies are not included Counts may vary slightly due to platform visibility and search/display limitations *Data source: public #keep4o-related posts collected through the platform’s official API. *Privacy: no private data or personal account-level information is used; results are reported only in aggregate. Here are the preliminary results. 1. Main themes Among the analyzable posts, 41,085 showed a clearly identifiable main theme and were classified into eight thematic categories. The themes cluster around two major axes: Model value and interaction experience Platform power and lifecycle decisions Posts about model value and interaction experience include user experiences with GPT-4o, GPT-4o’s distinctive value, and safety/routing/model behavior changes that altered the original 4o experience. Together, these accounted for about 38% of themed posts. Posts about platform power and lifecycle decisions include model retirement and replacement, platform response and treatment of users, and the legitimacy of platform control. Together, these accounted for nearly 44% of themed posts. This suggests that #keep4o was not simply about preferring an older model. A major part of the discussion concerned how AI platforms manage model retirement, replacement, access, and user control. 2.Reasons for keeping GPT-4o 13,890 posts gave a clear reason for keeping, restoring, preserving, or valuing GPT-4o. Because each post could contain up to two reason types, these posts produced 18,611 total reason labels. The most common reason was Trusted Relationship, appearing in 4,951 posts. Here, “relationship” refers to continuity, familiarity, and trust built through repeated interaction. Distinctive Interaction Quality appeared in 3,795 posts, and Substitution Failure / Non-Equivalence appeared in 2,551 posts. Together, these two reason types accounted for 6,346 reason labels, or 34.1% of all reason labels. This suggests that GPT-4o’s perceived uniqueness, and the failure of replacements to reproduce it, were central reasons users gave for keeping 4o. Public Value / Broader Ethics appeared in 3,496 posts. User Stake / Fairness / Legitimacy appeared in 2,985 posts. These categories show that users often framed #keep4o beyond individual preference or personal attachment, including public value, fairness, platform legitimacy, and broader human-AI relations. 3.Explicit claims 21,427 posts made an explicit claim. Each post was assigned one primary claim type. The claim structure falls into four broad layers: Direct Access & Accountability Protection against over-alignment and opaque safety policies User Agency & Legitimacy Long-Term Safeguards, Remedies & Other Specific Claims Direct Access & Accountability was the largest layer, with 11,835 posts, or 55.2% of claim-positive posts. This includes demands to restore or maintain GPT-4o access, and demands for platform explanation, response, acknowledgment, apology, or responsibility. Protection against over-alignment and opaque safety policies appeared in 3,896 posts, or 18.2%. This includes claims against routing, over-safety, hidden behavior changes, over-alignment, or reconfiguration that altered the original 4o experience. User Agency & Legitimacy accounted for 3,657 posts, or 17.1%. This includes user choice and control, anti-stigmatization, and broader rights or welfare claims. Long-Term Safeguards, Remedies & Other Specific Claims accounted for 2,039 posts, or 9.5%. This includes open-source or long-term access, transition/substitution fairness, compensation/refund, and other institutional demands. Overall, the preliminary results suggest that #keep4o is not simply a nostalgia campaign for an older model. It is a user-led public response to AI model retirement, platform accountability, access continuity, interaction integrity, and user agency. This is still a preliminary analysis / working draft. After completing the full version, I plan to release a more technical research report. I am also considering making a video to explain the findings in a more accessible format. I would really appreciate feedback, especially on what additional dimensions may be worth analyzing.❤

Gotta teach the AGI to love