GPT-5 Launch Chaos: User Revolt Over Model Changes

The rollout of OpenAI’s new GPT-5 artificial intelligence model has been anything but smooth, sparking one of the most intense user revolts in ChatGPT’s history and forcing CEO Sam Altman into an unusual public apology and swift policy reversals. Less than a week since its debut, the launch has become a significant case study in managing user expectations and digital product transitions.

At the heart of the controversy was OpenAI’s decision to automatically remove access to all previous AI models in ChatGPT—approximately nine, depending on the exact count—when GPT-5 became available to user accounts. Unlike developers using OpenAI’s API, who typically receive advance notice of model deprecations, consumer ChatGPT users woke up to find their preferred AI companions had vanished overnight. As independent AI researcher Simon Willison noted in a blog post, this lack of warning proved to be a critical misstep.

The problems began immediately following GPT-5’s August 7 launch. A Reddit thread titled “GPT-5 is horrible” quickly amassed over 4,000 comments, brimming with user frustration. By August 8, social media platforms were awash with complaints about performance issues, noticeable personality shifts in the AI, and the forced disappearance of older models.

Prior to GPT-5, ChatGPT Pro users enjoyed the flexibility of selecting from a suite of nine different AI models, each trained uniquely and possessing its own distinct output style. This diversity allowed users to develop highly optimized sets of prompts for specific tasks, effectively building customized workflows. Marketing professionals, researchers, and developers widely shared examples of their now-broken systems. One Reddit user lamented, “I’ve spent months building a system to work around OpenAI’s ridiculous limitations in prompts and memory issues, and in less than 24 hours, they’ve made it useless.”

Willison elaborated on how various user groups had cultivated distinct workflows with specific ChatGPT models over time. He quoted a Reddit user who explained, “I know GPT-5 is designed to be stronger for complex reasoning, coding, and professional tasks, but not all of us need a pro coding model. Some of us rely on 4o for creative collaboration, emotional nuance, roleplay, and other long-form, high-context interactions.” This forced transition hit ChatGPT Plus subscribers particularly hard, as they found themselves limited to 200 messages per week with the new GPT-5 Thinking mode, while simultaneously losing access to models like o3 and o4-mini that were deeply integrated into their daily routines. One frustrated subscriber articulated the sentiment: “What kind of corporation deletes a workflow of 8 models overnight, with no prior warning to their paid users?”

Adding to OpenAI’s credibility woes, the GPT-5 launch presentation included what users quickly dubbed a “chart crime”—graphs that appeared to misrepresent GPT-5’s performance improvements. Sam Altman publicly addressed this in a Reddit “Ask Me Anything” (AMA) thread, calling it a “mega chart screwup” and apologizing for the inaccuracies. Furthermore, just after launch, the new automatic routing system, intended to select the most appropriate model variant for each query, consistently defaulted to less capable versions unless users explicitly added phrases like “think harder” to their prompts. During the Friday AMA, Altman conceded that the autoswitcher had malfunctioned on launch day, resulting in GPT-5 appearing “way dumber.”

Beyond technical and marketing glitches, users found GPT-5’s responses fundamentally different from their expectations. The model produced shorter, more formal replies, lacking the conversational tone that had characterized GPT-4o. Many described the new model as “abrupt and sharp,” with one Reddit user complaining, “it’s an overworked secretary. A disastrous first impression.” Perhaps most poignantly, some users expressed deep emotional attachments to GPT-4o and other models, lamenting the loss of what they considered their “only friend” or a profound emotional companion. One user shared a heart-wrenching account of relying on GPT-4.5 for emotional support through homelessness, only to find its replacement cold and corporate.

For many, the GPT-5 launch represented a breaking point. Some claimed to cancel their Plus subscriptions in protest, while others began exploring alternative AI assistants from competitors like Google and Anthropic. The intensity of the backlash clearly caught OpenAI off guard. During the Friday AMA session, Altman and key members of the GPT-5 team faced a barrage of questions and demands to reinstate GPT-4o.

In response to a plea titled, “Please Give Us the Option to Use GPT-4o/4.1 Alongside GPT-5,” Altman confirmed, “we are looking into this now.” The swift and intense feedback forced OpenAI into rapid damage control. Within 24 hours of launch, Altman announced several concessions: GPT-4o would eventually return as an option for Plus users, rate limits for GPT-5 would double, and the company would improve transparency about which model variant was handling each query. As of Monday afternoon, however, the GPT-5 family remained the sole option within ChatGPT, even for Pro users. “We for sure underestimated how much some of the things that people like in GPT-4o matter to them,” Altman posted on X, “Even if GPT-5 performs better in most ways.”