OpenAI announced late Friday that it is updating its flagship GPT-5 model to make interactions feel “warmer and friendlier.” The change comes just weeks after the highly anticipated release, which CEO Sam Altman admitted was “a little more bumpy than we’d hoped for.” Some users had even voiced a preference for the older GPT-4o.
The company says the tweaks are “subtle” but should make GPT-5 more approachable. According to OpenAI, users can expect small touches like responses beginning with “Good question” or “Great start” not as flattery, but as a way to make the AI feel more natural. Internal tests reportedly show no increase in sycophancy compared to earlier versions.
At a dinner with journalists earlier this week, OpenAI executives tried to shift the conversation toward long-term plans beyond GPT-5. Still, the rocky launch remained a central topic. VP Nick Turley described GPT-5’s personality at launch as “very to the point,” while confirming that the newly announced update is designed to add warmth without losing precision.
With this adjustment, OpenAI hopes to address user feedback and strike a balance between accuracy and approachability in its most advanced model yet.
