OpenAI walks a tricky tightrope with GPT-5.1’s eight new personalities (arstechnica.com)

OpenAI released GPT-5.1 Instant and GPT-5.1 Thinking in ChatGPT, introducing eight preset “personality” styles (Professional, Friendly, Candid, Quirky, Efficient, Cynical, Nerdy, and Default) and claiming improvements in conversational tone, instruction-following, and benchmark performance. The launch comes amid legal and regulatory scrutiny and user complaints about prior model behavior, so OpenAI is emphasizing configurable communication styles while stressing that the presets simply alter prompt instructions — the underlying model capabilities remain the same. Technically, GPT-5.1 Instant is the faster default for most tasks, while GPT-5.1 Thinking simulates deeper reasoning for complex problems; both reportedly outperform GPT-5 on technical benchmarks like AIME 2025 and Codeforces. Instant also uses “adaptive reasoning” to decide when to spend more compute on a prompt, trading latency for accuracy when needed. OpenAI will roll the models out to paid subscribers first, then free users, and add them to the API (gpt-5.1-chat-latest for Instant; GPT-5.1 for Thinking) with adaptive reasoning enabled. Legacy GPT-5 versions remain available to paid users for three months. The update aims to balance user customization, performance gains, and safety/regulatory concerns, while leaving core model behavior and governance challenges largely unchanged.