GPT-5 gets a warmer, friendlier tone after a rocky rollout

OpenAI has started rolling out a personality tweak to its latest flagship model, aiming to make GPT-5 feel warmer, friendlier and more approachable in everyday use.

The refresh follows a launch that some users found rough around the edges, with a subset saying they preferred the conversational style of GPT-4o. This update is meant to address that feedback without changing the model’s core capabilities or inflating its praise.

The company describes the changes as subtle, adding small, genuine acknowledgments that set a positive tone—think brief nods like “Good question” or “Great start,” rather than over-the-top compliments. According to internal evaluations, the adjustment did not increase sycophancy compared to the previous GPT-5 personality.

At a recent off-the-record media dinner, leadership tried to steer discussion toward what comes after GPT-5, but the choppy rollout remained the backdrop. One executive summarized the issue this way: the initial GPT-5 came across as a bit too curt and “to the point,” and this refresh is designed to add a touch of warmth without slowing users down.

Why it matters: Tone shapes trust and usability. If the update lands as intended, conversations should feel more welcoming while staying concise—something power users and newcomers alike have been asking for. Expect further fine-tuning as more feedback rolls in.