OpenAI GPT-5 model faces backlash, company brings back legacy models for customers

In a swift reaction to user dissatisfaction, OpenAI has restored access to the GPT-4o model for its paid ChatGPT subscribers, just one day after the introduction of the much-anticipated GPT-5. This decision comes in response to considerable user criticism regarding the abrupt removal of GPT-4o, a model that many had relied upon in their daily tasks. Despite OpenAI's assertion that GPT-5 is its most sophisticated large language model yet, the rollout was met with a blend of enthusiasm and skepticism. During an AMA session on Reddit, CEO Sam Altman faced feedback from users who described GPT-5 as lacking the engaging personality of its predecessor, GPT-4o. One user candidly remarked that while GPT-4o felt like a conversational partner, GPT-5's responses seemed more mechanical and less personal. Moreover, several users reported issues with OpenAI's new real-time router, which was designed to optimize the selection of models for specific tasks. Altman acknowledged these concerns, noting that the router had not performed as expected and promised that GPT-5 would exhibit improved intelligence starting that day. He also assured users that they would soon be able to see which model is providing responses. Initially, the update had also removed the model selection feature in ChatGPT, which allowed users to choose from different versions, including GPT-4o. However, ChatGPT Pro, Team, and Enterprise users can now access GPT-4o again through the settings by selecting the “show legacy models” option, while free-tier users remain restricted to GPT-5. Altman stated that OpenAI would evaluate user engagement before determining the future support for older models and hinted at the potential return of GPT-4.1 if there is sufficient demand from Plus subscribers. He emphasized the company's commitment to stability and ongoing user feedback.