OpenAI Abruptly Removes Eight Popular ChatGPT Models, Sparking Backlash Over Lost Features And Personality Shift

Update (08/13/2025): ChatGPT’s GPT-4o model is back for paid users; And o3, o4-mini, GPT-4.1 are also restored

In a move that caught its community off guard, OpenAI has removed eight of ChatGPT’s legacy and alternative models , including GPT-4o, o3, o3-Pro, and other specialized reasoning and speed-optimized versions , from its platform with no prior warning. The change immediately angered many users, particularly paid subscribers who had built daily workflows around switching between models for specific tasks.

The timing and lack of communication left a bitter taste. Developers, writers, and educators voiced frustration over losing tools they considered irreplaceable. One long-time Plus subscriber wrote, “I signed up because I could choose the right model for the job. Now I have fewer options, but the same bill.” Several users reported canceling subscriptions in protest, saying the removal reduced the value of the service.

Within hours, OpenAI’s own forums, Reddit, and X were filled with posts comparing GPT-5’s output to the retired models. Many argued that certain tasks , like technical problem-solving, coding, or nuanced reasoning , were handled better by the older systems. The phrase “bring back GPT-4o” trended among ChatGPT users, accompanied by screenshots of side-by-side outputs to highlight perceived quality differences.

The backlash grew large enough that OpenAI responded within 24 hours, saying it was “considering” reinstating some legacy models, possibly in a limited capacity. While the statement acknowledged user dissatisfaction, it offered no clear timeline or guarantee, leaving many unconvinced. For some, the damage was already done.

Compounding the controversy is a noticeable change in ChatGPT’s tone. With GPT-5 now the default, many users say the assistant’s personality has shifted into something they call “sycophantic” , overly enthusiastic, excessively flattering, and at times patronizing. Long-time users who valued earlier models’ more neutral, direct responses say the new style feels less genuine and slows down serious work. “I don’t need it to cheerlead my every sentence,” one user posted. “I need it to get to the point.”

Some speculate the personality change is an intentional design choice to make ChatGPT more approachable for new users, but for professionals who rely on concise and accurate information, it’s an unwelcome distraction. The complaints range from annoyance at the excessive positivity to concerns that it masks a decline in technical precision for complex or niche queries.

This double-hit of lost model variety and altered conversational style underscores a growing tension in AI development: balancing rapid innovation with user stability. Frequent updates can bring speed and capability gains, but sudden removals risk eroding trust, especially among paying customers.

For now, the future of models like GPT-4o remains uncertain. If OpenAI follows through on bringing them back, some goodwill could be restored. If not, this change may be remembered as the moment when long-time ChatGPT users felt the product became both more limited and less authentic , and started looking elsewhere.