November 13, 2025 OpenAI has released GPT-5.1, an update to its main language model that introduces two new variants, tone-controlled responses and improvements in reasoning and clarity. The rollout began Nov. 12 for paid users, with access for free users to follow.
The update adds two models aimed at different types of work. GPT-5.1 Instant is designed for quick, conversational responses and focuses on warmer, more natural wording. GPT-5.1 Thinking is built for complex analysis and allocates more processing time for multi-step tasks such as coding, research and mathematical reasoning.
A new tone-selection feature lets users choose from preset conversational styles, including Professional, Friendly, Quirky and Cynical. OpenAI says the system adapts its language and explanations to match the selected tone, reducing the amount of manual editing required for different audiences.
The company reports improved benchmark results across coding, logic and mathematics. It also says the model produces fewer generic responses and follows instructions more consistently. Paid subscribers will retain access to the earlier GPT-5 model for three months during a transition period.
A simple example highlights the change in tone control and clarity. When asked to explain a neural network in one sentence, a previous version of the model gave a technical description referencing biological structures and interconnected nodes. GPT-5.1, when set to Professional tone, described it as a computing system that learns patterns from data. In Friendly tone, it called it a set of connected layers that learn patterns the more they are used.
Although GPT-5.1 remains part of the GPT-5 family, the update signals a shift toward models that are not only more capable but also more adaptable to different communication styles. OpenAI is continuing the rollout as the new versions replace the older defaults.
