OpenAI launched GPT-5 just a week ago, touting it as a standardized model based on an intelligent "router" that automatically selects the best style of answer, aiming to eliminate the need for manual model selection. However, the company soon added the form picker back, with three new modes: "Auto," "Quick," and "Reflect," with the ability to revert back to previous models such as GPT-4o, GPT-4.1, and o3 for paid subscribers.
Sam Altman, CEO, acknowledged that GPT-5 will get a personality update to become friendlier and less annoying, with a focus on personalization. The return of the model picker sparked widespread controversy, especially after performance issues on launch day prompted Altman to react to the criticism in a Reddit Q&A session.
Nick Turley, vice president of ChatGPT, admitted that the first experience wasn't perfect, but praised the team's speed in making improvements. Experts believe that the real challenge for OpenAI lies in finding the balance between speed of response and quality of answers, especially since many users have become emotionally attached to specific models, as seen in a symbolic "funeral" event for a model from a rival company.
These developments suggest that the "standardized model" may risk becoming a more complex experience if users' expectations are not met accurately.
Can OpenAI improve on its shortcomings and overcome this issue?
"GPT-5... from standardization to flop in a week!"
