Don't like ChatGPT's tone? Change it. That's the gist of OpenAI's latest update to the language model behind the industry-leading chatbot. GPT-5.1, announced in a blog post Wednesday, is supposed to be "more conversational" than its predecessor, and you have more flexibility to adjust its "personality" to suit your desires.
The new model is actually two models: GPT-5.1 Instant and GPT-5.1 Thinking. The former is the latest version of OpenAI's basic language model. The company said it's "warmer, more intelligent, and better at following your instructions" than its predecessor.
GPT-5.1 Thinking is the new iteration of OpenAI's advanced reasoning model. The company said its changes for Thinking focused on making it easier to understand. In its blog post, OpenAI reported that it's also more efficient -- generating fewer tokens on easier tasks while spending more time on difficult tasks. A reasoning model works by running a variety of operations to handle more complicated requests, so spending more time (and tokens) can result in better answers. (Disclosure: Ziff Davis, CNET's parent company, in April filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.)
You don't necessarily have to choose between the models. GPT-5.1 Auto can route simple requests to the Instant model and complex ones to the Thinking model.
OpenAI said it plans to roll out the changes gradually to paid users first, followed by free users. If you still want to use GPT-5, don't worry just yet: They'll still be available for paid users under the "legacy models" dropdown for three months.
Watch this: Parental Control for ChatGPT, AI Tilly Norwood Stuns Hollywood, Digital Safety for Halloween Night | Tech Today
03:08
Choose your ChatGPT
You've been able to customize the way the language model communicates with you for a while, with OpenAI including a series of presets that you could choose by going into settings. Those settings are changing and getting easier to find, the company said. The new choices are Default, Friendly, Efficient, Professional, Candid and Quirky. Friendly and Efficient are updated versions of the former Listener and Robot personalities. Professional, Candid and Quirky are new. The former Cynic and Nerd personas remain, as Cynical and Nerdy. Beyond those presets, you can change the chatbot's tone directly in your personalization settings -- including telling it to use fewer (or more) emoji.
"Instead of trying to build one perfect experience that fits everyone (which would be impossible), we want ChatGPT to feel like yours and work with you in the way that suits you best," Fidji Simo, OpenAI's CEO of applications, said in a blog post.
Read more: Learning the 'Personality' of an AI Video or Image Model Is the New Creative Work Hack
The tone of a chatbot matters, and one that is too friendly or reassuring can become sycophantic, posing serious mental health and safety concerns. Sycophancy caused OpenAI to roll back some updates to a ChatGPT model earlier this year.
"The best people in our lives are the ones who listen and adapt, but also challenge us and help us grow," Simo wrote. "The same should be true for AI."
English (US) ·