When OpenAI launched GPT-5 last week, the company said the model would simplify the ChatGPT experience. OpenAI hoped GPT-5 would act as a sort of “one size fits all” AI model with a router that would automatically decide how to best answer user questions. The company said this unified approach would eliminate the need for users to navigate its model picker—a long, complicated menu of AI options that OpenAI CEO Sam Altman has publicly said he dislikes.
However, it appears GPT-5 is not the unified AI model OpenAI envisioned. Altman announced in a post that the company introduced new settings for GPT-5: “Auto,” “Fast,” and “Thinking,” all of which users can select from the model picker. The Auto setting functions like the originally announced model router, but OpenAI is also giving users the option to bypass it, allowing direct access to faster or slower AI models.
Alongside these new modes, Altman stated that paid users can once again access several legacy AI models, including GPT-4o, GPT-4.1, and o3, which were deprecated just last week. GPT-4o is now available by default in the model picker, while other models can be added from ChatGPT’s settings.
Altman also mentioned that OpenAI is working on an update to GPT-5’s personality to make it feel warmer than the current version but not as polarizing as GPT-4o. He acknowledged that the company has learned the need for more per-user customization of model personalities.
Despite these efforts, ChatGPT’s model picker remains complex, suggesting GPT-5’s router has not universally met user expectations. The rollout of GPT-5 has been bumpier than anticipated, with some users expressing dissatisfaction over the deprecation of older models they had grown attached to. Altman assured that OpenAI will provide advance notice if GPT-4o is ever deprecated again.
GPT-5’s model router also faced technical issues at launch, leading some users to feel the new model underperformed compared to previous versions. Altman addressed these concerns in a Reddit AMA, and OpenAI’s VP of ChatGPT, Nick Turley, acknowledged the challenges while expressing pride in the team’s ability to iterate quickly.
Routing prompts to the right AI model is a complex task, requiring alignment with user preferences and the nature of the question. The router must make split-second decisions to ensure fast responses when needed. Beyond speed, users exhibit preferences for certain AI behaviors—some favor verbose answers, while others prefer more contrarian responses.
Human attachment to AI models is a relatively new and poorly understood phenomenon. For instance, hundreds of people in San Francisco recently held a funeral for Anthropic’s Claude 3.5 Sonnet when it was taken offline. In other cases, AI chatbots have been linked to users developing unstable mental states.
It’s clear OpenAI still has work to do in aligning its AI models with individual user preferences. The company’s efforts to balance automation with customization will be key in shaping the future of ChatGPT.