OpenAI’s GPT-5.1 brings smarter reasoning and warmer conversational styles
According to OpenAI, the new GPT-5.1 models are smarter, blending advanced adaptive reasoning with warmer, customisable personalities, making the AI enjoyable to talk to.
ChatGPT maker OpenAI has introduced a significant update to its large language models with the release of the GPT-5.1 series, comprising GPT-5.1 Instant and GPT-5.1 Thinking.
This upgrade focuses on enhanced capability and a more refined, enjoyable communication style.
“It’s a nice upgrade. I particularly like the improvements in instruction following, and the adaptive thinking. The intelligence and style improvements are good too,” Sam Altman, CEO of OpenAI, wrote on X.
The widely used GPT-5.1 Instant model is said to be “warmer by default and more conversational”. The model can exhibit playfulness while remaining both clear and useful.
A key functional improvement is enhanced instruction following, allowing the model to answer the question posed in a more reliable manner. For example, when directed to respond solely in six words, GPT-5.1 Instant adhered to the question consistently, offering a concise travel suggestion, according to an OpenAI blog post.
Furthermore, GPT-5.1 Instant now employs adaptive reasoning, allowing it to “think before responding” to complex queries for the first time, said the company. This mechanism contributes to more thorough and accurate answers and is reflected in “significant improvements on math and coding evaluations," it added.
Meanwhile, the advanced reasoning model, GPT-5.1 Thinking, has also been upgraded to be more efficient and easier to understand in everyday applications. This model adapts its thinking time dynamically, dedicating more time to difficult problems while speeding up responses to simpler ones. Its approach means less waiting time for straightforward requests and more comprehensive answers for complex tasks.
The responses generated by GPT-5.1 Thinking are clearer, featuring “less jargon and fewer undefined terms”. This makes the model more approachable for explaining technical concepts, such as complex sports statistics.
Additionally, GPT-5.1 Thinking adopts a default tone that is both warmer and more empathetic. When addressing user distress over a minor mishap, the model offered immediate reassurance, stating, “A coffee spill reads as ‘busy human,’ not 'idiot’”. The model framed the user's action as resilience, noting, “You’re a person who had a chaotic moment, wiped it up, and still showed up."
The release included efforts to allow users to customise ChatGPT’s tone and style. The idea is to make tailoring the conversational tone effortless. Building on earlier personalisation features, the preset options have been refined. Existing settings such as default, friendly, and efficient remain, and new options have been added, including professional, candid, and quirky. These personality settings apply consistently across all models.
For deeper control, OpenAI is experimenting with the ability for users to tune specific characteristics directly, such as the required conciseness, warmth, or scannability of the responses, or the frequency of emoji use.
The updated GPT-5.1 models are also better at honouring existing custom instructions, providing greater control over behaviour and tone, OpenAI noted.
The rollout began immediately with paid users, including Pro, Plus, Go, and Business subscribers. To facilitate transition, the older GPT-5 models will remain available for three months in the legacy models dropdown for paid subscribers. Earlier this month, OpenAI gave Indian users free ChatGPT Go access for a year.
In recent months, Anthropic and Google have also introduced significant updates to their AI models, focusing on enhanced reasoning, improved speed, multi-modal capabilities, and advanced agentic features.
For example, Anthropic released Claude Sonnet 4.5 in September 2025, with large leaps in reasoning, math and coding performance, extended autonomous task horizons, and top-tier agentic/computer-use abilities.
Earlier this year, Google DeepMind rolled out updates to the Gemini 2.5 Flash and Flash-Lite models, featuring better instruction-following, reduced output token costs, improved multimodal understanding, and stronger agentic tool use.
Edited by Swetha Kannan


