OpenAI Launches the most powerful, intelligent, and conversational version of its AI (again): this is GPT-5.1

Lisa Simpson had up to five different cats, four of them named Snowball (each with its respective number). However, when we think of Lisa’s cat, only the name “Snowball” comes to mind, and all those cats seem to be unified in a single representation. I wonder if that is what happens to most people with ChatGPT.

OpenAI has just introduced GPT-5.1, the new and improved ChatGPT model. Once again, it is more intelligent, reliable, and conversational than its long list of predecessors (GPT-5, GPT-4.1, GPT-4.5, GPT-4.1-mini, and many others).

GPT-5.1 in ChatGPT is rolling out to all users this week.

It is smarter, more reliable, and a lot more conversational.https://t.co/SA1Q1GPyxV

— OpenAI (@OpenAI) November 12, 2025

The technology company has announced that this model has already begun to roll out to all its users, starting with paying subscribers, with the intention that it will fully replace GPT-5 within three months. Until that time, if you have a paid plan, you will be able to select either GPT-5 or GPT-5.1 from the model dropdown menu and compare their performances.

“The sunsetting period for GPT-5 does not affect the availability of other legacy models. Going forward, when we introduce new ChatGPT models, our approach is to provide users with sufficient time to assess the changes and share their feedback, which will enable us to continue innovating our cutting-edge models during a seamless transition. Sunsetting periods will be clearly communicated with ample notice,” OpenAI explained.

GPT-5.1 Instant and GPT-5.1 Thinking

As with previous releases, OpenAI has developed two versions of this new AI: one intended for daily use (GPT‑5.1 Instant) and another designed for deeper queries requiring more extensive reasoning (GPT‑5.1 Thinking).

The company has designed GPT‑5.1 Instant to be more intelligent, but also friendlier, providing helpful and clear responses with a personable tone. Furthermore, the model’s capacity for following instructions and adapting to user inquiries has been improved.

Another important innovation is that, for the first time, an Instant model can utilize adaptive reasoning. Thus, as explained by OpenAI, it can: “decide when to think before responding to more complex questions, resulting in more comprehensive and accurate answers without sacrificing speed. This is reflected in significant improvements in mathematical and programming assessments such as AIME 2025 and Codeforces.”

For its part, GPT-5.1 Thinking is an advanced reasoning model that has been improved in key areas such as its ability to explain complex ideas in simple terms and its capacity to adapt the processing time devoted to each task based on its level of complexity. In addition, the company also aimed to increase the level of empathy and approachability of this model when responding to queries.

Likewise, those who are uncertain about which version to use at any given moment or who, simply, do not wish to take the time to decide, may select the GPT-5.1 Auto mode in the ChatGPT model selector. In this way, the AI will determine whether to use Instant or Thinking after assessing the characteristics of the query.

ChatGPT returns to being compliant

With this latest release, OpenAI has placed considerable (indeed, significant) emphasis on GPT-5.1’s capacity to be kind, helpful, and even flattering. This adjustment comes in response to the wave of negative reactions generated by the personality change in GPT-5 relative to previous models, as it became more direct and less complimentary.

“Users clearly told us that a good AI should not only be intelligent but also pleasant to interact with. GPT-5.1 significantly improves both intelligence and communication style.” In short, it is evident that people prefer a more accommodating attitude over a neutral tone.

Because of this, in addition to making GPT-5.1 more personable, the company has also fine-tuned its settings to facilitate the customization of ChatGPT’s tone and personality. To do so, you simply need to click on your profile, select the “Personalization” section, and choose one of the eight options (Default, Professional, Kind, Sincere, Quirky, Efficient, Nerdy, and Cynical). When you select one of these personalities, it will be applied to all ChatGPT models that you use within your account.

On the other hand, OpenAI is also experimenting with additional customization possibilities within the “Personalization” section, providing users with greater freedom and capability to configure their AI. Another option that may also be effective is guiding the chatbot in a conversation and specifying your preferences.

This shift in OpenAI’s approach to the personality of its AI models is rather striking when we consider that the company is eager to emphasize its commitment to users’ mental health and well-being. ChatGPT is a tool, not a friend, and permitting this boundary to be crossed, placing the responsibility in how each person configures their chatbot, entails risks of behavioral disorders that could ultimately pose a genuine social problem.

Photo: Gemini