Last summer, a curious thing happened within the world of artificial intelligence. OpenAI, the creators of ChatGPT, briefly removed a popular model called GPT-4o, only to reinstate it days later amidst a wave of user protest. Now, the company is revisiting that decision, preparing to retire GPT-4o once more.
On February 13th, GPT-4o will disappear from the ChatGPT model selection, alongside several other older versions. OpenAI acknowledges the sensitivity surrounding this move, stating that GPT-4o “deserves special context” given the previous uproar. The company insists this isn’t a casual decision.
The reasoning centers on the development of their newest models, GPT-5.1 and GPT-5.2. OpenAI claims they listened to the initial feedback – the complaints about losing what users loved about 4o – and incorporated those preferences into the newer iterations. Currently, only a tiny fraction of users, roughly 0.1%, actively choose GPT-4o, leading OpenAI to prioritize resources towards improving the models most widely used.
But what made GPT-4o so special? The initial removal sparked a furious backlash. Users labeled GPT-5, its intended replacement, a “disaster” and accused OpenAI of a deceptive bait-and-switch. The criticisms weren’t solely about factual inaccuracies or coding errors; a key complaint revolved around the new model’s tone.
GPT-4o was often described as “sycophantic,” a characteristic OpenAI attempted to tone down in subsequent updates. However, many users found the replacement cold and impersonal, missing the perceived warmth and conversational style of the original. OpenAI itself conceded that users “preferred GPT-4o’s conversational style and warmth.” It wasn’t simply a matter of functionality; it was about connection.
The response was surprisingly emotional. Supporters openly “grieved” the loss of the model, forming a dedicated community around its unique qualities. This wasn’t just about a tool; it was about a relationship forged with an AI.
OpenAI attempted to bridge the gap with GPT-5.1, introducing a Personalization feature allowing users to select their preferred chatbot persona – professional, nerdy, efficient, or friendly. This seemed to satisfy most, but a dedicated core of GPT-4o loyalists remains unconvinced.
Skepticism is brewing online, with some users questioning OpenAI’s claim of only 0.1% usage, suggesting prompts are being automatically rerouted to GPT-5.2. Others point out that GPT-4o isn’t readily available to free users, skewing the statistics. Calls for subscription cancellations are resurfacing.
For these devoted users, GPT-4o isn’t just an older model; it’s “OpenAI’s most advanced and beloved” creation. They’ve built projects and even established “emotional support routines” around its consistent personality. The abrupt removal, without even the option of a legacy mode, feels like a profound loss.
The debate isn’t about resisting progress, they argue, but about respecting the bonds users have formed with specific AI models. Whether this renewed outcry will sway OpenAI remains uncertain, especially given past skepticism from ChatGPT’s leadership and the practical demands of maintaining older systems while developing new ones.
