OpenAI says GPT-5 should be more friendly now

GPT-5 hasn't even been out for two weeks and OpenAI has already had to tinker with its tone.

This Tweet is currently unavailable. It might be loading or has been removed.

In a post on X over the weekend, OpenAI confirmed that it was making GPT-5 "warmer and friendlier" after initial complaints that it was too stodgy and cold in its interactions with users. Apparently, users should keep an eye out for phrases like "good question" in response to prompts, though OpenAI was quick to note that this shouldn't include outright flattery or sycophancy.

Power users of ChatGPT have had myriad issues with the new model since its launch earlier this month, with its supposed unfriendliness frequently coming up in social media posts. It also proved problematic for Sam Altman and company that OpenAI took away the option to manually switch to older models, though the option was eventually restored for paid users.

Cursory searches on places like X and Reddit reveal that there doesn't seem to be much of a change in how people are talking about GPT-5 yet. Many of the same complaints are still coming up, so it might be a while before OpenAI fully fixes the problem. And, as Reddit user MysticalMarsupial smartly put it, it's possible that OpenAI wants GPT-5 to be a bit colder emotionally for a very good and practical reason.

"Lawsuits. It's lawsuits," the user said. "They are afraid that you guys will fry yourselves into psychosis and that your families will sue OpenAI."