OpenAI revealed late on Friday plans to update its latest model to make it “warmer and friendlier.”
The company recently unveiled the eagerly awaited GPT-5, a launch that CEO Sam Altman conceded was “a little more bumpy than we’d hoped for.” Some users expressed a preference for the preceding model, GPT-4o.
OpenAI aims to tackle some criticisms with this update, asserting that the changes are “subtle” yet will render GPT-5 “more approachable now.”
The company stated in a social media post, “You’ll notice small, genuine touches like ‘Good question’ or ‘Great start,’ not flattery.” They also indicated that “Internal tests show no rise in sycophancy compared to the previous GPT-5 personality.”
During a dinner event for journalists, OpenAI executives aimed to highlight the company’s future endeavors beyond GPT-5, although, as Max Zeff notes, the challenging launch was the elephant in the room. In terms of model friendliness, VP Nick Turley remarked that GPT-5 was “just very to the point,” but the newly announced update would make it feel warmer.