Sam Altman brings GPT-4o back after GPT-5 outrage, but most won't get it
OpenAI’s recent rollout of GPT-5, its latest and most advanced large language model, has been met with significant user backlash, rather than the anticipated acclaim. Despite being made available to hundreds of millions worldwide, both free and paid ChatGPT subscribers have voiced strong dissatisfaction, many claiming the new model is a downgrade from its predecessor, GPT-4o, and lamenting the abrupt removal of older versions.
The widespread discontent came to a head during a Reddit Ask Me Anything (AMA) session with OpenAI CEO Sam Altman. Users flooded the forum with complaints, asserting that GPT-5 felt “dumber” and was disrupting workflows that had been meticulously built around GPT-4o. Beyond performance concerns, many users expressed frustration over GPT-5’s perceived “colder” personality and the company’s decision to suddenly unify its offerings under a single new model, leaving them no option to revert to their preferred, older iterations.
Responding to the torrent of feedback, Altman acknowledged the issues, stating that the OpenAI team had heard the users’ concerns. He announced a partial concession: Plus subscribers would regain the option to continue using GPT-4o, with the company monitoring usage to determine the duration of this support. Altman also shed light on a technical hiccup within GPT-5 itself. The model incorporates a new real-time router designed to select the most appropriate underlying model for a given task. However, Altman admitted that this router had not been functioning as intended, contributing to the perceived decline in response quality. He assured users that improvements were being implemented, promising that GPT-5 would “seem smarter starting today” and that OpenAI would enhance transparency regarding which model was answering specific queries.
While the company has moved to address some of the grievances, access to legacy models remains tiered. ChatGPT Pro, Team, and Enterprise users can now switch back to GPT-4o via a “show legacy models” banner in their app settings. However, free users currently have no such option, remaining locked into the new GPT-5 experience. When questioned about the potential return of even older models like GPT-4.1, Altman indicated that this would only occur if there was sufficient demand from Plus users. Concluding the AMA, Altman reiterated OpenAI’s commitment to stabilizing the models and continuously incorporating user feedback. This episode highlights the delicate balance between rapid technological advancement and ensuring a seamless, satisfactory user experience for a vast and diverse global audience.