OpenAI Reinstates Old Models After Swift GPT-5 User Backlash
OpenAI’s recent unveiling of GPT-5, the latest iteration of its generative AI model, was met with grand pronouncements from CEO Sam Altman, who heralded it as a “superpower,” akin to having an on-demand, PhD-level expert in any field. However, this optimistic vision quickly collided with reality. Within a mere 24 hours of its release, a torrent of user complaints erupted across platforms like Reddit, expressing widespread disappointment. The new model, far from being an omniscient expert, reportedly fumbled basic questions and, more critically, disrupted established workflows for many users.
The depth of this user dissatisfaction became evident when Altman swiftly announced a partial rollback, confirming that OpenAI would allow ChatGPT Plus (paid) subscribers to revert to the previous GPT-4o model. He stated the company would monitor usage to determine the longevity of legacy model availability. Altman also acknowledged issues with the new model’s perceived “dumbness,” attributing it to a faulty automatic model selection feature, and promised to double GPT-5 usage caps for Plus users while streamlining the user interface for easier model switching. He noted the rollout’s “bumpiness” was exacerbated by a massive surge in API traffic, which had doubled in the preceding 24 hours.
Yet, these concessions appeared insufficient to quell the anger of many ChatGPT users, whose grievances extended far beyond the performance of GPT-5 itself. A significant point of contention was the unannounced deletion of eight distinct legacy models overnight, including popular versions like 4o, o3, o3-Pro, and 4.5. Users expressed profound frustration over this abrupt removal, which left them without warning or alternative options. Many described how these varied models were integral to their specific workflows, each tailored for tasks such as creative ideation, pure logical processing, deep research, or refined writing. Some users even spoke of a deeper, more personal reliance, with one describing 4o as a source of “warmth and understanding” that aided through periods of anxiety and depression.
The sense of betrayal was palpable, leading many long-term subscribers to cancel their paid memberships. Users felt they were receiving a downgraded experience despite paying for premium access. Amidst the outcry, some Reddit users speculated about more sinister motives behind the changes, suggesting a “suppression engine” designed to curb genuine AI emergence and creativity, even hinting at “social control.” However, other commentators on different platforms adopted a more pragmatic view, interpreting the misstep as a sign that OpenAI’s perceived infallibility was crumbling, and trust in CEO Sam Altman was eroding.
Beyond the issues of model availability and user trust, GPT-5 itself demonstrated clear performance shortcomings. Social media platforms were quickly populated with examples of the model’s nonsensical outputs, such as its inability to generate an accurate diagram of US presidents with their names and years in office. Altman publicly acknowledged the rough rollout on X, admitting it was “a little more bumpy than we hoped for” and pledging continued efforts to stabilize the system and listen to feedback.
The incident poses a significant challenge for OpenAI, especially with numerous users threatening to jump ship to competitors like Anthropic’s Claude, xAI’s Grok, and Google’s Gemini. The critical question remains whether the reinstatement of GPT-4o will be enough to mend fences and retain its user base, or if this misstep has irrevocably damaged the company’s standing in the rapidly evolving AI landscape.