GPT-5.1: A smarter, more conversational ChatGPT (openai.com)

OpenAI today upgraded the GPT‑5 family to GPT‑5.1, shipping two main variants: GPT‑5.1 Instant (the default, more conversational model) and GPT‑5.1 Thinking (the advanced reasoning model). Key technical improvements include stronger instruction following, “adaptive reasoning” in Instant (it decides when to pause and think for harder queries), and dynamic thinking-time allocation in Thinking (longer on complex tasks, faster on simple ones). Early benchmarks reportedly show notable gains on math and coding evaluations such as AIME 2025 and Codeforces. Responses are also designed to be clearer and less jargon-heavy, with warmer default tones and better persistence on complex problems. A system-card addendum will document safety updates. Rollout begins today to paid tiers (Pro/Plus/Go/Business) with a gradual public roll-out and API releases later this week (gpt-5.1-chat-latest for Instant; GPT‑5.1 for Thinking). GPT‑5 remains available under legacy models for three months; Auto routing will select the best GPT‑5.1 model for queries. The update also expands personalization: refreshed presets (Default, Friendly, Efficient, plus Professional, Candid, Quirky) and experimental fine-grained controls for concision, warmth, scannability, and emoji use that apply across ongoing chats. For the AI/ML community this blends improved frontier capability with UX and alignment work—adaptive reasoning and better instruction following ease integration into complex workflows, while personalization and clearer outputs improve human-AI interaction and deployment flexibility.