GPT-5 gets a warmer, friendlier tone after a bumpy debut

OpenAI is rolling out a refinements update to GPT-5 aimed at making conversations feel warmer and more welcoming, without changing how the model reasons or answers.

The launch of GPT-5 drew enormous attention but also mixed reactions, with some users saying they preferred the style of GPT-4o. Leadership has acknowledged that the debut didn’t go as smoothly as hoped.

To address those concerns, the latest tuning focuses on subtle cues that make the assistant more approachable. Expect brief, genuine acknowledgments such as “Good question” or “Great start” to help set a collaborative tone — the goal is encouragement, not flattery.

According to OpenAI, internal evaluations show no increase in sycophancy compared with the earlier GPT-5 personality, suggesting a friendlier voice without becoming overly agreeable.

Company executives have been keen to discuss what’s next beyond GPT-5, but the rocky kickoff remains a theme in conversations. As one leader put it, the original behavior was very to-the-point; this update softens the edges so interactions feel a bit warmer while staying concise.