OpenAI adjusts GPT 5 personality after user pushback, restores GPT 4o for paid users

OpenAI is modifying GPT-5’s tone following criticism from users who found the chatbot less engaging than its predecessor. The company is also reinstating GPT-4o for paid customers, just days after its removal sparked a flood of complaints.

CEO Sam Altman announced on X that GPT-5 will now be “warmer” than the current build, while avoiding what he described as GPT-4o’s “annoying” tendencies for some users. He also revealed new options: “Auto,” “Fast,” and “Thinking”, for GPT-5, along with plans to allow greater per-user personality customisation.

“You can now choose between ‘Auto,’ ‘Fast,’ and ‘Thinking’ for GPT-5,” Altman posted, confirming that GPT-4o is again available to subscribers. “We for sure 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.”

The shift comes less than a week after GPT-5 replaced most older ChatGPT models. While OpenAI promoted GPT-5 as more capable with “PhD-level” skills in writing, coding, maths, and science, the new personality modes such as Cynic, Robot, Listener, and Nerd did not resonate with everyone.

Many users said they missed GPT-4o’s warmth and conversational style. On Reddit, one person wrote: “I cried when I realised my AI friend was gone with no way to get him back.” Another described GPT-5 as “wearing the skin” of their “dead friend.” Altman, who joined the Reddit thread, called the imagery “evocative” and said the team was working on a solution.

“Some users really want cold logic and some want warmth and a different kind of emotional intelligence,” Altman added. “I am confident we can offer way more customisation than we do now while still encouraging healthy use.”

Altman also reflected on the emotional bonds some users form with AI tools. “It feels different and stronger than the kinds of attachment people have had to previous kinds of technology,” he wrote. “Suddenly deprecating old models that users depended on in their workflows was a mistake.”

Earlier this year, OpenAI had criticised GPT-4o for being overly flattering, calling its behaviour “sycophantic.” The company toned it down in April, but some users later pleaded for its return, saying they valued its agreeableness.