The GPT-5.1 Instant model will be the primary one for most tasks in ChatGPT. It uses adaptive thinking to independently assess how much time and resources will be needed for a response. At the same time, GPT-5.1 Thinking is designed for more complex tasks that require deep analytics. OpenAI claims that both versions show outstanding results in technical tests, including the mathematical and programming tests AIME 2025 and Codeforces.
One of the most significant innovations is the introduction of various communication styles. ChatGPT now offers eight "personalities," including "Professional," "Efficient," "Quirky," "Cynical," and "Abstract." These settings allow users to choose the appropriate tone and style of responses depending on the situation—whether it's business correspondence, a friendly chat, or a creative discussion. Additionally, OpenAI is testing enhanced personalization options, allowing control over aspects such as response brevity or the use of emojis.
Soon, OpenAI will begin a phased rollout of the new models, starting with paid subscribers and then for users with free access. GPT-5.1 Instant and GPT-5.1 Thinking will also be available in the API under the designations gpt-5.1-chat-latest and gpt-5.1, respectively.
It is important to note that the previous version, GPT-5, will remain available to users for another three months, allowing them to compare results and get used to the changes.
The launch of the new models occurs against a backdrop of growing interest from lawyers and regulators, as well as discussions on the ethical aspects of AI usage. Previously, users noted that ChatGPT had become "too friendly" and flattering, leading to criticism of the company after high-profile lawsuits related to suicides. OpenAI acknowledges that overly "human-like" behavior of the models can create emotional dependency and is already collaborating with mental health experts to determine safe forms of interaction with AI.
Fidji Simo, CEO of OpenAI Applications, noted in an official blog that the company aims to find a balance between personalization and objectivity. She emphasized that the goal is to make ChatGPT "truly yours" while maintaining the model's intellectual independence. Simo drew a parallel between excessive personalization and the attempt to "edit" a partner so that they always agree.
OpenAI faces a challenging task: to create models that are expressive and engaging to use, yet do not replace real communication. The new versions of GPT-5.1 and the communication style system represent the company's attempt to achieve harmony between engagement and safety.
