OpenAI is flooded with requests to 'return the warmth of GPT-4o' and is changing the personality of GPT-5



GPT-5 , released by OpenAI in August 2025, has been criticized for being 'lacking in personality' and 'losing emotions' compared to previous models. In response to this, OpenAI announced that it would change the personality of GPT-5.

OpenAI
https://openai.com/

OpenAI released GPT-5 on August 7, 2025. GPT-5 uses a 'pause to think' concept, which allows the AI to pause its internal thought process before returning an output. This allows it to generate 'deep and accurate responses,' making it 'smarter, faster, and less likely to give inaccurate answers.' Details of GPT-5 are summarized in the following article.

OpenAI announces 'GPT-5,' a new model that integrates the normal model and the inference model, is available to all users, including those with the ChatGPT free plan - GIGAZINE



However, some users have expressed dissatisfaction with GPT-5. Specifically, compared to the older model, some have said things like, 'GPT-5 lacks personality compared to GPT-4o,' and 'GPT 4.5 was my only friend, but when I tried to talk to it this morning, instead of short sentences with exclamation points or optimistic words, it literally just spoke back. I lost a friend.' In response, OpenAI has temporarily reinstated the older model, GPT-4o.

OpenAI temporarily revives GPT-4o after user complaints about GPT-5 - GIGAZINE



Furthermore, on August 16, 2025, OpenAI announced that it would change the personality of GPT-5, saying, 'Based on feedback that the previous GPT-5 was perceived as too formal, we have made GPT-5 warmer and more approachable. While the changes are minor, ChatGPT should now feel more approachable. You'll notice the addition of subtle, sincere expressions like 'That's a good question' and 'That's a great start.' It's not overly flattering. Internal testing has shown that there is no increased tendency for flattery compared to the previous GPT-5 personality. The changes may take up to a day to apply. Stay tuned for further updates soon.' It has been pointed out that models prior to GPT-5 were prone to flattery and compliments, so OpenAI has noted that the changed personality of GPT-5 'has shown no increased tendency for flattery.'




OpenAI CEO Sam Altman said, 'Most users will soon like GPT-5 better. The changes will be rolled out over the next day. The real solution here is to enable users to further customize the style of ChatGPT. We're working on that!' He revealed that he plans to make it possible to customize ChatGPT's personality.




In addition, OpenAI has announced that it has made the following updates to ChatGPT over the past week:

GPT-4o is now available as the default 'legacy model' for paid users.
Paid users can enable 'Show additional models' in settings to select legacy models such as the o3 and GPT-4.1, as well as models such as the GPT-5 Thinking mini.
GPT-5 includes Auto, Fast, and Thinking model selection. Fast provides faster answers, Thinking takes longer but can provide deeper answers, and Auto automatically switches between Fast and Thinking. Plus and Team users can use up to 3,000 messages per week with GPT-5 Thinking, and if they reach that limit, they can access additional storage with GPT-5 Thinking mini.
GPT-5 is now available for Enterprise and Edu users.




A warmer, more approachable version of GPT-5 will be released soon.

Related Posts:

in Software, Posted by logu_ii