OpenAI has announced the release of a new version of its flagship series of language models — GPT-5.1, which includes two main models: GPT-5.1 Instant and GPT-5.1 Thinking.
OpenAI has announced the release of a new version of its flagship series of language models — GPT-5.1, which includes two main models: GPT-5.1 Instant and GPT-5.1 Thinking.
According to the official description, GPT-5.1 improves both the communication style («warmer, more „human“ interaction») and intellectual capabilities (better execution of instructions, adaptation to the complexity of tasks).
GPT-5.1 is an attempt to combine technical power and better UX/interaction style. The company is betting not only on «mind», but also on «communication». The emergence of two models with specialization (Instant vs Thinking) is a signal that AI platforms are moving from «one model for everything» to more flexible architectures for different types of tasks. For developers and users, this means: the ability to choose the appropriate mode, adapt the response style, improve interaction.
The developers note that GPT-5.1 Instant: is designed for the most common scenarios that require a quick response and a more natural, «conversational» style. OpenAI notes: «Now warmer, more intelligent, and better at following your instructions.»
GPT-5.1 Thinking, in turn, is focused on more complex queries — the model adapts response time depending on the task: simple queries are solved quickly, and complex ones — with deeper analysis.
Both models get new tone settings: users can choose exactly how the model should sound — for example, friendly, professional, humorous style.
Older models (including GPT-5) will remain available some time for comparison so that users can gradually transition.
OpenAI has published a System Card Addendum for GPT-5.1, which describes updated safety metrics and new assessments in areas such as mental health and emotional dependence.
The rollout began on November 12, 2025 — initially for paid ChatGPT users (Plus, Pro, Business), and later — for all subscribers and users of the free version.
Also, access to new models is provided through the API: Instant is designated as gpt-5.1-instant, Thinking is designated as gpt-5.1-thinking.
OpenAI admits that the previous model GPT-5 received mixed reviews: users complained about a «cold», less human tone, and awkward interaction.
If you have already used GPT-5, you can compare the results with the new version. In particular, whether the model has really become «warmer» or understands instructions better.
For organizations using ChatGPT or OpenAI API, it is important to test whether the new version meets the requirements (quality, speed, tone) and whether it will affect existing integrations.
The new system card indicates that OpenAI is paying more attention to users' mental health and the potential psychological effects of interacting with AI.
OpenAI did not release detailed technical benchmarks (such as new performance metrics or model parameters) at the time of the announcement. Users and experts are watching for real-world results: whether the problem-solving performance will improve or the «humanness» of the interface will improve.
