OpenAI launches GPT-5.1: a warmer, smarter ChatGPT that finally listens better

Three months after the debut of GPT-5, OpenAI has rolled out GPT-5.1, a refined version of its flagship model that promises more natural conversations, smarter reasoning, and better control over tone. The update marks a major usability leap for ChatGPT, addressing long-standing complaints about robotic replies and poor instruction-following.

OpenAI says GPT-5.1 brings improvements to both its Instant and Thinking models—the two modes that power ChatGPT’s everyday interactions. The Instant model, which handles the majority of user prompts, has been fine-tuned to sound “warmer, more intelligent, and better at following your instructions.” The result is a version of ChatGPT that feels more like a human conversation partner and less like a data parser.

The update also introduces new “personality” options, letting users choose how the model sounds. Options include default, friendly, efficient, professional, candid, and quirky—giving more control over tone than ever before. OpenAI has long struggled to strike the right balance between personality and precision, and GPT-5.1 seems to be its most flexible take yet.

A big win for users is the model’s improved ability to follow specific instructions. In tests shared by OpenAI, GPT-5.1 correctly obeyed prompts such as “reply in six words,” a task earlier versions often failed. This makes the model far more reliable for structured writing, coding, or educational use.