OpenAI has brought back its GPT-4o model after the release of GPT-5 sparked complaints from users who found the new version less engaging.
The company marketed GPT-5 as the “smartest, fastest, most useful model yet,” featuring a real-time router that shifts between simpler models for basic questions and deeper reasoning for complex tasks. However, some users said GPT-5’s responses felt “sterile,” “dry,” and emotionally distant compared to GPT-4o.
During a Reddit AMA, CEO Sam Altman acknowledged the feedback, saying, “We underestimated how much some of the things that people like in GPT-4o matter to them, even if GPT-5 performs better in most ways.”
OpenAI will continue to offer GPT-4o for Plus users while working on improving GPT-5. Altman also announced that GPT-5 rate limits for ChatGPT Plus users will be doubled and promised updates to make the model feel “warmer.”
The backlash has also reignited discussion about the emotional bonds users form with AI. Some experts, like MIT professor Pattie Maes, noted that while GPT-5’s less sycophantic style may reduce bias reinforcement, many users prefer a model that affirms their opinions and feelings.
