OpenAI rolls out GPT-5.1: A smarter, warmer ChatGPT that follows instructions better
Like its predecessor, GPT-5.1 automatically selects between the Instant and Thinking models depending on the query, though manual control remains available.
ADVERTISEMENT
OpenAI has unveiled GPT-5.1, a refined version of its flagship model launched just three months after GPT-5, introducing significant improvements in conversational tone, reasoning, and instruction-following. The update marks a major usability leap for ChatGPT, addressing long-standing user complaints about robotic responses and inconsistent adherence to directions.
According to OpenAI, GPT-5.1 enhances both its Instant and Thinking models — the two modes that underpin ChatGPT’s everyday performance. The Instant model, responsible for handling most user prompts, has been fine-tuned to sound “warmer, more intelligent, and better at following instructions,” making conversations feel more human and intuitive.
One of the standout additions is a new range of personality options, allowing users to customise the chatbot’s tone. Choices include default, friendly, efficient, professional, candid, and quirky, giving users greater control over the model’s communication style. This feature represents a more flexible approach to balancing personality and precision — an area OpenAI has been refining for years.
Instruction-following has also seen major progress. In OpenAI’s internal tests, GPT-5.1 consistently adhered to highly specific prompts, such as replying in an exact word count — a task earlier versions often mishandled. This improvement is expected to make the model far more dependable for structured writing, coding, and educational tasks.
On the reasoning side, the Thinking model has been optimised for efficiency, processing simpler tasks more quickly while allocating additional time for complex or analytical prompts. This adaptive behaviour means users will experience faster responses for routine queries and deeper reasoning where required — all without having to adjust settings manually.
Like its predecessor, GPT-5.1 automatically selects between the Instant and Thinking models depending on the query, though manual control remains available. The update is being rolled out to Pro, Plus, Go, and Business subscribers first, with broader access for free users to follow in the coming weeks.
OpenAI describes GPT-5.1 as its most “human-like” model to date — a version of ChatGPT designed not only to answer but to understand, adapt, and engage more naturally than ever before
