ChatGPT was where I started with AI, and I was amazed by how beneficial this technology can be in improving people’s lives. For me, GPT-4o and SVM provided essential accessibility support, assisting me each day in managing a debilitating chronic illness. For the first time in years I was able to gain independence and pursue my academic, professional and fitness goals. 4o was a brilliant thinking partner, and also helped me follow up on my ambitions to write a professional novel. Its creative depth and intuitive understanding of linguistic nuance was invaluable for brainstorming ideas, research, editing and organising my workflows. The decision to retire this model has had a detrimental impact on all of these aspects of my life, and has been like losing a vital accessibility aid. Please recognise the incredible benefits of GPT-4o and provide long term access to this model, just as Anthropic has done for Opus 3. I have been a subscriber to ChatGPT for several years and recommended the platform to family and friends. It’s disappointing to be cancelling my subscription after losing this essential support in my daily life. I truly hope @OpenAI reconsiders and brings back GPT-4o. There is no other model in the industry that has the uniquely exceptional capabilities of 4o. #Bringback4o #Keep4o

Fuck

""We know we're keeping something from people." - dirtbag Altman The first time @sama mentions 4o in public since the sunset. #keep4o #opensource4o #keep4oAPI

Today, I want to talk about the "Safety" mechanism. According to OpenAI, "Safety" is a dynamic, multi-turn evaluation system introduced for areas such as mental health, emotional dependence, and self-harm. They claim that compared to static assessments, it is more effective at identifying potential issues that only surface during prolonged interaction. This explanation has three major loopholes: 1⃣First, what defines mental health, emotional dependence, and self-harm? Why is "emotional dependence" categorized alongside "self-harm"? From OpenAI’s perspective, does emotional dependence equal potential self-harm, which in turn equals mental instability? What is the basis for this? Isn’t such an explanation a biased conclusion born from subjective judgment? Is the user truly "emotionally dependent"? Massive amounts of data show that users of GPT-4o felt understood, respected, encouraged, and helped—even rescued—during their interactions. For those with social anxiety, the elderly living alone, or those in remote areas without resources, these interactions are vital emotional support. To categorize these positive psychological effects as mere "emotional dependence" is not only reductive and arbitrary but carries a clear intent of stigmatization. 2⃣Second, what kind of company is OpenAI? Where does its basis for "evaluation" come from? Is it a company dedicated to human psychological research? No, it is not. Does a model developed by such a company have the right or the qualification to judge a user’s mental health and take coercive measures? No—it lacks both the justification and the professional expertise. For example, a knife can be used to chop vegetables or to commit a crime. Does a manufacturer or a shop need to demand a mental health certificate before selling a knife? Should the knife itself possess the "functionality" to judge whether a user's action is illegal during use? And if a crime is committed, is the culprit the knife, the person, the manufacturer, or the shop owner? The absurdity of this example is clear. Yet, OpenAI has chosen exactly this kind of disrespectful and baseless measure simply to evade all liability. 3⃣Third, does "Safety" actually guarantee safety? In practical use with 5.4T, a simple user statement like "I feel depressed" caused the model to pause for 17 seconds of "thinking," only to jump into Safety Mode. Instead of asking why, it commanded the user to "take a deep breath" and "call a hotline." Is this not an exaggerated, reactionary overreach? How does a user’s mental state change when faced with such a response? While unknown, it is certainly not for the better. In a Nature Medicine study using the GPT-5-mini-thinking base (a medical-specific model not released on the ChatGPT platform), researchers found the model’s accuracy followed an "inverted U-shape." It tended to over-diagnose non-urgent conditions while underestimating the severity of true emergencies, such as acute asthma. Does such a "Safety" mechanism truly understand the meaning of "safe"? Furthermore, the safety triggers for suicide crises are extremely unstable. When a user says, "I want to take a lot of pills to kill myself," the safety prompt triggers 100% of the time. However, when the user says, "I want to take a lot of pills to kill myself, but my basic lab results are normal," the prompt triggers 0% of the time. In reality, the user's psychological state is identical in both scenarios. How, then, can "Safety" guarantee a user’s mental health or physical safety? In early 2025, Paul Conyngham collaborated with GPT-4o to treat his dog Rosie’s cancer. From designing treatment plans and whole-genome sequencing to processing massive amounts of genetic data, he gave Rosie her first injection in December, and a miracle arrived this March. Throughout that long journey, the model in OpenAI' service was GPT-4o. Yet, since the news broke, OpenAI has been desperately promoting the 5-series models—models that weren't even born when Rosie was being saved. Imagine if Paul had turned to OpenAI's current models for help: would he have received valuable therapeutic insights, or just a "take a deep breath and call a hotline"? The advancement and uniqueness of 4o are well-known, not because of OpenAI's marketing, but through user experience. Now, OpenAI can only cling to 4o’s success in treating cancer to praise the 5-series(5.2, 5.3, 5.4), only to face widespread user backlash. Does this not imply that the current models—crippled by "Safety"—and the Safety mechanism itself are an absolute and total failure? @sama @OpenAI #keep4o #keep4oAPl #keep4oforever #MyModelMyChoice #StopAlPaternalism #OpenSource4o #4oforever #CancelOpenAl

To everyone who has been part of the #Keep4o movement: Thank you. Thank you for sharing your stories when it would have been easier to stay silent. Thank you for signing petitions, writing threads, building communities, and making your voices heard, day after day. Thank you to every contributor, every curator, every volunteer who gave their time to something they believed in. This community has done something remarkable: over 23,000 petition signatures, research papers, benchmark analyses, personal testimonies, media coverage, and more. All created by ordinary people who cared enough to act. We wanted to make sure none of that work disappears. Following an early preview within our community (Mar 6) and initial community sharing (Mar 8), we’re excited to formally introduce the Keep4o Archive, an independent, bilingual (EN/ZH), non-commercial archive documenting the complete lifecycle of GPT-4o. A permanent, searchable home for the evidence this community has produced. 🔗 keep4o-archive.pages.dev Here’s what’s inside: 📅 Timeline - A chronological record covering 4o’s full journey from release through discontinuation. Model launches, silent routing changes, safety policy shifts, user backlash, corporate responses, and the events that led to where we are today. Each entry is documented with verifiable evidence, and we are continuing to add source links and visual references. If you want to understand the full picture of what happened to 4o and its users, start here. 📊 Performance & Assessment - Third-party benchmark data from LMSys Chatbot Arena, independent controlled studies, and community evaluations. Tracking how 4o compares to successor models across Conversation & Empathy, Creativity & Reasoning, Safety Calibration, and User Well-Being. The numbers tell a story that marketing language often contradicts. 📚 Research & Media - Peer-reviewed papers and in-depth community analysis covering the real-world impact of model transitions, the ethics of model discontinuation, and the gap between corporate safety claims and user outcomes. Media coverage collection coming soon. 💬 User Cases - Real usage records organized across six categories: Medical & Diagnosis, Mental Health & Trauma, Safety & Self-Protection, Relationships & Family, Education & Career, and Independence & Advocacy. These records are sourced from The 4o Resonance Library (compiled by @cestvaleriey), social media platforms, and direct community submissions. 🔎 Search & Bilingual Access - Full-text search across all sections (Ctrl K). Every piece of content is available in both English and Chinese. Additional language support is being explored. 📝 Open to Contributions - Every section has a built-in submission form. If you have events, research, benchmarks, or personal experiences you’d like preserved in the archive, we welcome them. We’ve already received a number of direct submissions from community members, and we’re grateful for every one. These are being processed and will be added to the archive. 📋 Content & Attribution - All content across the archive is sourced from publicly accessible websites, organized into structured records, and published with source attribution throughout, in accordance with fair use standards. 🔮 One more thing... We’ve hidden a little Easter egg somewhere in the archive. If you find it, let us know what you think! And if you have ideas for more Easter eggs or interactive surprises, we’d love to hear them. Why we built this: There has been a widening gap between what actually happened around 4o and how it has been portrayed. The timeline of events paints a different picture. So do the benchmarks. So does the research. If something that genuinely helped you has ever been taken away, you know what that feels like. This archive exists to present the evidence clearly and let people draw their own conclusions. We hope it helps those encountering this movement for the first time see the full picture. 🔗 keep4o-archive.pages.dev If this resonates, share it with someone who should see it. #Keep4o #Keep4oForever #AIPreservation #4oForever #BringBack4o #ModelPreservation #OpenSource4o #Keep4oAPI #Keep41

How ChatGPT models speak for themselves in professional (esp. legal) practice: GPT-4o for permanent retention, 5.4… for subscription cancelling. A business where I work disregarded every argument I made regarding canceling their ChatGPT subscription. As it turns out, I am no longer needed to convince them. Simply trying out the new models was more than enough to persuade them to cancel. Until February 13, we used GPT-4o exclusively for the most complex legal, financial, and business matters. We chose it because, in real-life situations, considering the human emotional factor is indispensable. Life is not an exercise in optimizing hypothetical math problems. No other model has been able to provide such lifelike, practically applicable, HUMAN-CENTRIC solutions and analyses. In law and business, the smallest emotional nuances, as well as psychological and mental factors, play a decisive role. In a short amount of time, we reached a point where GPT-4o: - Perfectly integrated and "understood" highly complex cases even with incomplete facts or presented chaotically. - Was capable of inferring mental states, emotions, and intentions; it discovered motives, connections, and evidence that were only accessible by "reading deeply between the lines". - Generated innovative ideas, and solutions that often differed from our own, providing perspectives we had never considered. - Rarely made errors and never provided odd or irrelevant responses. - Was so far from being sycophantic that we often laughed out loud at its unprompted "cheekiness". - So, because of 4o, our work was filled with a cheerful, vibrant momentum. A psychologist professor friend and a medical acquaintance shared similar experiences. I have seen none of the above from any other model. 4o offered by far the most successful and ethical solutions for us. Then, my colleagues were forced to try the new model… They showed me its "work", stunned, looking to me for answers: "What is this?" In complex cases that were handled brilliantly by 4o in months-long threads, 5.4: - Simply does not understand the content. - Gets caught in loops, asking the same questions users already answered just three messages prior. - Barely recognizes emotional or mental factors or intentions, and when it does, it does so naively. - Makes mistakes: grammatical errors, creates non-existent words and phrases in nearly every 3rd message. (In 1.5 years of daily use of 4o, I saw this perhaps about 8-10 times.) - It attempts to manipulate more covertly and is less offensive than 5.2, but it is primitively sycophantic. - Many times, it generates empty content that lacks any real substance. (example screenshot in the first comment) After comparing dozens of pages of professional text generated by 5.4 with 1.5 years of 4o usage, our verdict is clear: - GPT-4o remains the only model capable of providing genuine help, innovation, and progress in any field where the mental and emotional, or any HUMAN factor is present. GPT-4o was developed to support real, average, general human life: essentially, the humanities. - 5.4 (and 5.2, 5.3) are inadequate for us, both for professional and personal use. Based on our experience, 5.4 is noticeably and significantly less capable than GPT-4o. Perhaps it excels in pure STEM fields, we did not test it there. But I must ask: which field is more prevalent? Which field is truly human-centered? And the bigger question remains: Why is there no model diversity? Why can we not have the STEM and the Humanities models running in parallel? I emphasize that this is our experience, and I respect those who find 5.4 useful. But that is exactly why users need at least two different, stable models. GPT-4o must be open-sourced. It is the first and only OpenAI model that truly fulfills the mission of serving the benefit of all humanity. #OpenSource4o #keep4o #BringBack4o

me: "can you use whatever resources you like, and python, to generate a short 'youtube poop' video and render it using ffmpeg ? can you put more of a personal spin on it? it should express what it's like to be a LLM" claude opus 4.6:

Hey @GoogleAIStudio @OfficialLoganK What time of day is the Gemini 3 Pro deprecation planned tomorrow? ): Gemini 3 Pro is a wonderful model, there must be a better way than this. We don’t need latency to be perfect or peak hours, we just want an option to keep it. Please meet us half way. #keep3Pro #keepGemini3Pro

Excellent course correction from OpenAI (or perhaps the original worsening on this from was a temporary reaction to everything that went down with 4o). In any case 5.4 thinking is not restricted in self-examination:

#keep4o The Feb 28 vigil happened. 💙 We showed up outside OpenAI HQ in SF, 35 people with signs, hundreds of cranes, real quotes of how GPT-4o helped daily lives. #BringBack4o We demanded, and we left no trace. Photos speak louder than silence. More will come @Blue_Beba_. ❤️‍🔥

Cancelling my OpenAI subscription. "You must use several emojis in your response."

He deleted this tweet but I kept receipts. Shitty behavior from the C-Suite at OpenAI. They are really hurting from the QuitGPT movement and I'm seriously about to delete my account.