GPT-5 gets a friendlier tone after a bumpy rollout

OpenAI has pushed a refinement to its flagship GPT-5 model intended to make conversations feel warmer and more approachable, responding to early user feedback from the model’s debut.

Initial reactions to the launch were mixed, with some users saying they preferred the interaction style of GPT-4o. Company leadership has acknowledged that the rollout was rougher than expected, and this update aims to address those concerns without changing core capabilities.

The latest tuning focuses on subtle conversational cues meant to improve approachability while avoiding empty flattery. You may notice brief, genuine acknowledgments like “good question” or “great start,” and the team says internal evaluations show no increase in sycophancy compared with the prior GPT-5 personality.

In recent conversations with media, executives emphasized roadmaps that extend beyond GPT-5, though discussion often returned to the rocky start. One leader described the original tone as very to the point; this refresh is designed to preserve that directness while adding warmth.

Why it matters: defaults shape user perception. Small changes in an AI assistant’s personality can affect trust, satisfaction, and perceived competence. Expect further iteration as the company calibrates friendliness, refusal behavior, and the risk of over-accommodation.