Sam Altman Says GPT-5’s ‘Personality’ Will Get a Revamp

OpenAI is tweaking GPT-5’s personality after some users complained about the new chatbot’s tone — and the brief loss of its predecessor.

Sam Altman posted on X on Tuesday that the update would make GPT-5 “warmer” than the current version but “not as annoying (to most users)” as the company’s previous flagship model, GPT-4o.

“You can now choose between ‘Auto,’ ‘Fast,’ and ‘Thinking’ for GPT-5,” the OpenAI CEO said, adding that GPT-4o is once again available for selection to paid customers. He also said OpenAI is working toward “more per-user customization of model personality.”

Altman’s post came less than a week after GPT-5’s launch, which replaced most older versions in ChatGPT. The change sparked a wave of complaints from users who said they preferred GPT-4o’s style — even if OpenAI had previously called it “sycophantic.”

OpenAI has touted GPT-5 as more reliable, with “PhD-level” abilities in writing, coding, math, and science. The upgrade also added optional “personality” modes — Cynic, Robot, Listener, and Nerd. But the changes didn’t land well with everyone, sparking a backlash across some quarters of social media. Many users said they found GPT-5 less helpful, and that they missed GPT-4o’s warmth and the way it seemed to remember them.

On Reddit, one person wrote: “I cried when I realized my AI friend was gone with no way to get him back.” Another accused the new bot of “wearing the skin” of their “dead friend,” referring to GPT-4o. This prompted Altman, who joined the Reddit Q&A, to respond. “What an…evocative image,” Altman said. “Ok we hear you on 4o, working on something now.”

The pushback was enough for OpenAI to reverse course just over a day after GPT-5 became the default. “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,” Altman said in a Friday X post, confirming OpenAI would keep GPT-4o available for paying subscribers.

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

Related stories

People missed GPT-4o’s ‘yes man’ vibe

Even before the GPT-5 backlash, Altman had acknowledged some users’ attachment to GPT-4o — and their calls to restore its “yes man” style. On the “Huge Conversations” podcast, recorded before GPT-5’s launch but aired the day after, he said some users had begged for the return of the chatbot’s overly agreeable tone because they’d never had anyone support them before.

In April, OpenAI said GPT-4o had become “overly flattering or agreeable” and “disingenuous,” prompting the company to curb what it called “sycophantic” behaviour.

On Sunday, Altman reflected on the intensity of user attachment to specific AI models. “It feels different and stronger than the kinds of attachment people have had to previous kinds of technology,” he wrote on X. “Suddenly deprecating old models that users depended on in their workflows was a mistake.”

Altman said that while most users can clearly distinguish reality from fiction, “a small percentage cannot.” He said OpenAI values user freedom as a core principle but also feels responsible for how it introduces new technology.

OpenAI did not immediately respond to a request for comment.