OpenAI rolls out GPT‑5.1 with warmer conversation and adaptive reasoning
OpenAI announced GPT‑5.1 on November 12, 2025, updating the existing GPT‑5 family with two refined models and expanded personalization controls. The release focuses on improving conversational tone and reasoning behavior while making it easier for users to shape how ChatGPT responds.
Two refreshed models: Instant and Thinking
The GPT‑5.1 release includes two main variants:
-
GPT‑5.1 Instant — positioned as the default, most-used model. It is warmer by default, more conversational, and better at following instructions. The model also uses adaptive reasoning in some cases to decide when to allocate extra processing to produce a more considered reply while preserving responsiveness on routine queries.
-
GPT‑5.1 Thinking — the advanced reasoning variant. It allocates thinking time dynamically, spending less time on simple requests and more on complex problems. Responses are described as clearer and less jargony, with improvements on math and coding benchmarks cited (AIME 2025 and Codeforces).
Both models support the same adaptive reasoning capability, allowing Instant to “think before responding” for harder questions and Thinking to scale effort more precisely by task complexity.
Personalization: more tones and finer control
Personalization options have been expanded so ChatGPT can match different conversational styles more readily. Existing presets remain and new ones have been added: Professional, Candid, and Quirky, alongside Default, Friendly, Efficient, Nerdy, and Cynical. These base styles apply across all models.
In addition to presets, OpenAI is experimenting with more granular tuning from personalization settings, including sliders for conciseness, warmth, scannability, and emoji frequency. These adjustments are intended to apply immediately across ongoing conversations, and the system can proactively suggest preference changes during chats when it detects recurring tone requests.
Rollout, API names, and transition details
- Rollout timing: GPT‑5.1 Instant and GPT‑5.1 Thinking began rolling out on November 12, 2025, starting with paid tiers (Pro, Plus, Go, Business) and then expanding to free and logged-out users over the following days to keep performance stable.
- Enterprise and Edu: enterprise and education customers receive a seven-day early‑access toggle, off by default.
- API access: both models are scheduled to appear in the API later in the week — GPT‑5.1 Instant will be exposed as gpt-5.1-chat-latest, and GPT‑5.1 Thinking as GPT‑5.1, both with adaptive reasoning.
- Transition period: GPT‑5 (Instant and Thinking) will remain available in ChatGPT under the legacy models dropdown for paid subscribers for three months to allow comparison and adaptation. GPT‑5 Pro will be updated to GPT‑5.1 Pro soon.
OpenAI also notes that GPT‑5.1 Auto will continue routing queries to whichever model is best suited, so manual selection is not required in most cases.
Safety and documentation
OpenAI points to a system card addendum for more information on safety approaches specific to GPT‑5.1: system card addendum.
For full details, see the original announcement on OpenAI’s site: https://openai.com/index/gpt-5-1/
