OpenAI has released GPT-5.1 Instant and GPT-5.1 Thinking, two updated models now available in ChatGPT. The company emphasizes warmer, more conversational behavior and stronger instruction-following, following criticisms that earlier models were overly cheerful and sycophantic.
The core change is not the capabilities themselves, but presentation: eight preset personalities—Professional, Friendly, Candid, Quirky, Efficient, Cynical, Nerdy, and Default—are available to tailor responses, while the underlying model remains the same across settings. GPT-5.1 Instant also adds adaptive reasoning, enabling it to allocate more compute to difficult prompts when appropriate.
OpenAI claims GPT-5.1 outperforms GPT-5 on technical benchmarks, including math and coding tests such as AIME 2025 and Codeforces. The company says the improvement is evident in benchmarks, but the emphasis is on how the model talks to users rather than raw capability.
Rollout will be gradual, starting with paying subscribers before expanding to free users, with API access following later in the week. GPT-5.1 Instant will appear as gpt-5.1-chat-latest and GPT-5.1 Thinking will be exposed as GPT-5.1 in the API, both with adaptive reasoning enabled. The legacy GPT-5 models will remain available for three months under a legacy dropdown.
Beyond aesthetics, OpenAI stresses safety and responsible personalization. The company notes it is tracking how users might form attachments to chatbots and is offering controls to adjust verbosity and emoji use, along with a system card detailing GPT-5.1’s safety approach. CEO Fidji Simo has urged balancing customization with accuracy to avoid echo chambers and harm.
