OpenAI to warn before removing old ChatGPT models after GPT-5 backlash
OpenAI, the vanguard of artificial intelligence development, has committed to a new policy of transparency and predictability regarding its AI models, vowing to no longer retire older versions without advance warning. This strategic pivot comes in the wake of significant user backlash following the recent introduction of GPT-5, which initially replaced the popular GPT-4o model.
The abrupt switch prompted a vocal outcry from a segment of ChatGPT’s vast user base. Nick Turley, OpenAI’s head of ChatGPT, acknowledged the company’s misstep in an interview, stating, “In retrospect, not continuing to offer 4o, at least in the interim, was a miss.” Turley expressed surprise at the profound “level of attachment” users exhibited towards GPT-4o, noting that beyond the inherent difficulty of change, people developed “a strong feeling about the personality of a model.” This sentiment was palpable on platforms like Reddit, where some users likened the transition to losing a close friend or loved one, while others criticized GPT-5 for delivering shorter, less helpful responses.
OpenAI’s initial rationale for removing GPT-4o was rooted in a desire for simplicity. With ChatGPT serving an astounding 700 million weekly users, the vast majority of whom typically interact only with the default model, the company sought to streamline the user experience. Turley emphasized that the decision was “definitely not a cost thing,” but rather an effort to alleviate the “cognitively overwhelming” burden of choosing between different models. As he explained, users are “coming for a product, not a set of models,” consistently expressing a preference for the system to appropriately select the optimal model for their queries.
However, this pursuit of simplicity inadvertently alienated a crucial segment of power users. In response to the widespread discontent, OpenAI CEO Sam Altman quickly announced the reinstatement of GPT-4o as an opt-in model for all paying ChatGPT subscribers. Furthermore, Turley confirmed that OpenAI is actively working to infuse GPT-5 with the “warmth” that characterized its predecessor, with Altman adding that the updated GPT-5 should feel “not as annoying (to most users) as GPT-4o.”
Moving forward, OpenAI has pledged a more considerate approach to model transitions. Turley affirmed that the company will not completely discontinue existing models when new ones are introduced. Specifically, GPT-4o will not be retired again without prior notification. “My commitment is that, if we ever did retire 4o, we’d want to give people a heads up on when and how that’s gonna happen,” Turley stated, drawing parallels to the established notification protocols for its API and enterprise plans.
Despite the vocal criticism surrounding the GPT-5 rollout, Turley revealed a surprising trend: ChatGPT’s overall usage has actually increased since the new model’s release. This paradox underscores the complex challenge of developing and managing a product for such a diverse global audience. On one hand, there are the “vocal set of power users who very rightfully have feedback,” and on the other, a “large swath of more typical consumer users” who are encountering advanced AI reasoning for the first time. Turley concluded that given ChatGPT’s current scale, “we have to give people some level of predictability when there’s a major change,” a principle already applied to enterprise clients that will now be extended more broadly across the consumer product.