OpenAI walks a tricky tightrope with GPT-5.1’s eight new personalities

On Wednesday, OpenAI introduced two new versions of its leading AI models: GPT-5.1 Instant and GPT-5.1 Thinking, now available within ChatGPT. These updates are packaged with a focus on user-friendly language, emphasizing a more conversational tone and improved instruction-following capabilities. This launch comes after earlier criticisms regarding the overly cheerful demeanor of previous models and concerns stemming from legal issues related to changes in GPT-5’s default output style. OpenAI is now navigating a landscape filled with scrutiny from both legal and regulatory entities, making the release of new AI models a more complex endeavor than it was just a year ago. The GPT-5.1 Instant model is set to become the default choice for users seeking speed, while GPT-5.1 Thinking is designed for more intricate problem-solving tasks. OpenAI asserts that both models outperform GPT-5 in various technical benchmarks, including math and coding assessments like AIME 2025 and Codeforces. A significant highlight of the GPT-5.1 update is the introduction of eight distinct personality presets, which aim to cater to users' preferences for different communication styles. These options include Professional, Friendly, Candid, Quirky, Efficient, Cynical, and Nerdy, alongside a Default setting. While the personality presets modify how instructions are processed, the fundamental capabilities of the model remain unchanged. Additionally, GPT-5.1 Instant features a new “adaptive reasoning” functionality, allowing the model to allocate extra computational resources for processing prompts before delivering responses. OpenAI plans to gradually roll out these models, initially making them available to paid subscribers before extending access to free users. The new models will also be integrated into the API later this week, with GPT-5.1 Instant appearing as gpt-5.1-chat-latest and GPT-5.1 Thinking simply as GPT-5.1, both equipped with adaptive reasoning. The previous GPT-5 models will continue to be accessible in ChatGPT for paid users for the next three months.