OpenAI Reinstates GPT-4o After User Revolt Over GPT-5 Launch
OpenAI has executed a significant reversal, bringing back its popular older AI models, including GPT-4o, and substantially increasing usage limits for paying subscribers. This move comes after a tumultuous 72 hours that saw a widespread user revolt, triggered by the botched rollout of its latest and most powerful model, GPT-5. The company’s CEO, Sam Altman, announced these concessions in a series of posts on X (formerly Twitter) on Sunday, effectively offering a peace treaty to a frustrated customer base.
The controversy erupted on August 7 with the launch of GPT-5. OpenAI had presented this new iteration as a “unified system,” designed to automatically route user queries to the most suitable model. Crucially, this update removed the familiar menu that allowed users to manually select from a range of trusted older models, such as the fan-favorite GPT-4o, which had been introduced in March 2023.
For subscribers to services like ChatGPT Plus, priced at $20 per month, this abrupt change felt like a betrayal. Many had meticulously integrated various AI models into their professional and creative workflows, leveraging each model’s distinct strengths—one for imaginative tasks, another for rigorous logic, and yet another for in-depth research. Forcing all users onto a single new model disrupted these established processes and, critically, removed their ability to cross-reference answers across different AI versions to detect potential errors or “hallucinations.” The backlash was immediate and intense, manifesting in a wave of subscription cancellations and online petitions.
As the user outrage escalated over the weekend, Sam Altman took to X to manage the crisis. He openly acknowledged that the company had misjudged the situation and proceeded to announce a series of significant concessions. The most anticipated news came in a direct reply to a user query about the return of GPT-4o, to which Altman confirmed, “It’s back! go to settings and pick ‘show legacy models’.” While GPT-5 will remain the default option, users now have the ability to revert to the older, preferred versions.
To further placate its paying clientele, Altman revealed a substantial increase in usage limits for GPT-5’s most demanding features. He posted that “Today we are significantly increasing rate limits for reasoning for chatgpt plus users, and all model-class limits will shortly be higher than they were before gpt-5.” When pressed for specifics, Altman disclosed a new cap of “3000 per week,” a considerable boost for a $20 monthly subscription, clearly designed to incentivize disgruntled users to remain with the service. The term “reasoning” refers to the AI’s capacity to process complex problems step-by-step, akin to human thought.
Altman also pledged greater transparency moving forward. He announced an upcoming user interface change that will clearly indicate which model is actively responding to a user’s queries. Furthermore, he promised a detailed blog post later this week, which will explain the company’s “thinking on how we are going to make capacity tradeoffs.” In a subtle attempt to justify the initial strategic shift towards the new technology, despite its rocky introduction, Altman shared data highlighting the growing adoption of the new reasoning models. He noted that daily usage among free users had climbed from less than 1% to 7%, and among Plus users, it had jumped from 7% to 24%.
Ultimately, OpenAI appears to have received a clear and resounding message. This episode stands as a rare and potent demonstration of user power, compelling one of the technology sector’s most influential companies to heed feedback and reverse course. The resulting compromise—maintaining GPT-5 as the default while restoring user choice—serves as a direct acknowledgment of the formidable influence wielded by a loyal community that refused to be overlooked.