OpenAI Upgrades ChatGPT With GPT-5.1 AI Models, Brings Friendlier Conversations and Less Jargon

OpenAI introduced the GPT-5.1 family of artificial intelligence (AI) models on Wednesday. The first major update to the company's frontier models, released in August, brings improved tonality, intelligence, and instruction adherence. A total of three models have been released, GPT-5.1 Instant, GPT-5.1 Thinking, and GPT-5.1 Auto. The San Francisco-based AI giant highlighted that all of these models now respond to the user in a warmer and more empathetic manner. The upgrade will first be rolled out to the paid subscribers, but is expected to be made available in the free tier in the future as well.

OpenAI Introduces GPT-5.1

In a blog post, the AI giant detailed the newly released models. OpenAI said both GPT-5.1 Instant and Thinking have started rolling out to ChatGPT Go, Plus, Pro, and Business subscribers. Developers can find these models in the application programming interface (API) later this week, with Instant listed as GPT-5.1-chat-latest and Thinking listed as GPT-5.1.

More importantly, the company stated that GPT-5 will remain available in ChatGPT for three months under the legacy models drop-down, which is only available to paid subscribers. After that, the model will be sunset. However, the older models such as GPT-4o and GPT-4.1 will continue to be available.

Coming to the GPT-5.1 Instant, which will be the default model for the free tier after rollout, the company has added several new improvements. It is now said to be “warmer by default and more conversational.” OpenAI claimed that during testing, participants were surprised by the model's playfulness.

This improvement also addresses the biggest complaint with the fifth generation of GPT-based large language models. Many users had complained that the responses felt dry and robotic when compared to GPT-4.

Another area of improvement is instruction following. The model now follows the prompts closely, generating responses within the constraints of the request. Additionally, it now comes with adaptive reasoning, which allows the model to decide when to think before responding.

GPT-5.1 Thinking has also been improved. Its responses are now “more efficient and easier to understand.” The model uses less jargon and poorly defined words, and focuses on conveying the meaning. It can also automatically adapt its thinking time based on the complexity of the question.

Apart from this, OpenAI also introduced GPT-5.1 Auto, which is the improved auto router that automatically switches the models from Instant to Thinking based on the prompt.

Separately, the AI giant has also improved the personalisation tool in ChatGPT. The tool fine-tunes the responses of the chatbot based on presets. The Default mode has been improved, while the Listener and Robots will now be known as Friendly and Efficient, respectively.

Three new presets, Professional, Candid, and Quirky, have now been added. Professional is described as polished and precise, Candid is direct and encouraging, while Quirky is playful and imaginative. The Cynical and Nerdy presets remain unchanged.