OpenAI has released GPT-5.1, an update to its language model series, aimed at making interactions with its ChatGPT service more intelligent and conversational. According to the company’s announcement the update introduces two refined models, GPT-5.1 Instant and GPT-5.1 Thinking, which are rolling out to all users, starting with paid subscribers.
This release follows what the publication VentureBeat described as “mixed reviews” for the initial GPT-5, which some users found lacking in performance. The GPT-5.1 update is positioned as a significant improvement in both capability and user experience.
Improvements to the core models
OpenAI states that both models have been upgraded to have a “warmer” and more natural communication style.
- GPT-5.1 Instant: Described as the most-used model, it is now designed to be better at following instructions. For the first time, it uses “adaptive reasoning,” allowing it to spend more time on complex questions to provide more accurate answers while remaining fast on simple queries.
- GPT-5.1 Thinking: The advanced reasoning model is now more efficient. It dynamically adjusts its processing time, responding faster to simple tasks and more persistently to complex ones. OpenAI also notes that its responses contain less jargon, making it more approachable for explaining technical topics.
More personalization options
Alongside the model updates, OpenAI is expanding options for users to customize ChatGPT’s tone. New preset styles like “Professional,” “Candid,” and “Quirky” have been added to existing options. The company is also experimenting with more detailed controls, allowing users to adjust sliders for characteristics like conciseness or warmth. A key change is that any adjustments to style or custom instructions now apply immediately to all conversations, including ongoing ones.
The rollout of GPT-5.1 is gradual. Paid users get access first, followed by free users. For a transition period of three months, paid subscribers will still be able to select the older GPT-5 models. OpenAI stated this is to give people time to compare the versions and provide feedback. The company also confirmed that both new models will be made available to developers via its API.
Sources: OpenAI, VentureBeat
