ChatGPT 5.1 is smarter, nicer, and better at actually doing what you asked (www.techradar.com)

OpenAI has made GPT‑5.1 the new default in ChatGPT for paid users, packaging two tuned variants—GPT‑5.1 Instant (fast, default) and GPT‑5.1 Thinking (slower, deliberative)—and using adaptive reasoning to switch between them when a query is ambiguous or complex. Instant promises low latency without losing warmth or strict instruction-following, while Thinking has been reworked to avoid jargon and be more transparent about its steps. The company will keep older models available for at least three months and roll features out to free users later, signaling a phased, feedback-driven deployment and a commitment to iterative updates rather than waiting for GPT‑6. Alongside the model upgrade, ChatGPT gains real‑time personality controls (tone, emoji frequency, verbosity) that can be adjusted mid-conversation, letting users tune style and communicative behavior without restarting chats. For the AI/ML community this is significant: it’s a production example of model orchestration and adaptive inference at scale, plus stronger controllability and interpretability baked into UX. The release highlights practical trade-offs between speed, transparency, and user alignment, and creates a useful testbed for studying dynamic routing, human-in-the-loop preference shaping, and ongoing model iteration in deployed LLM systems.