OpenAI has released GPT-5.1 with two personalities because 800 million users do not want the same AI

OpenAI has launched GPT-5.1an update of its flagship model that comes in two variants:

1. Instant (conversational and “warmer”),
2. Thinking (deep reasoning).

The real novelty is not in the technical metrics, but in something more prosaic: you can now choose between eight conversation tones, from “professional” to “cynical.” It is recognizing that AI as a mass product needs segmentation. It is no longer enough to have just one assistant for everyone.

The new model selector for Plus users. Image: Xataka.

Why is it important. OpenAI has 800 million users with radically different expectations. Some want a neutral and efficient assistant. Others seek warmth and empathy. Some have even developed problematic emotional ties to the chatbot.

The company tries to solve this with personality adjustments, but the underlying problem remains: ChatGPT keeps pretending to be a person, a consistent entity that knows you. This generates the same risks of emotional dependence that have motivated mental health demands and alerts.

The facts:

• GPT-5.1 Instant improves in math and programming, and for the first time uses “adaptive reasoning”: decide when to think harder before answering.
• GPT-5.1 Thinking, for its part, dynamically adjusts its processing time according to the complexity of the question, being twice as fast in simple tasks and twice as slow in complex ones.

The eight available tones (Default, Professional, Friendly, Sincere, Quirky, Efficient, Geek, Cynical) work by injecting different instructions into each prompt. The capabilities of the model do not change, only the presentation changes.

Yes, but. The speed of the launch has come at a cost. OpenAI itself admits in its technical documentation that GPT-5.1 presents “known security regressions” compared to the October version. They prioritized time-to-market over exhaustive testing, something striking in a company under intense regulatory scrutiny due to cases of vulnerable users.

Furthermore, personalization has limits that OpenAI has had to explicitly acknowledge: “taken to the extreme, personalization would be useless if it only reinforced your worldview.” It is admitting that you are walking a tightrope between engagement commercial and social responsibility.

Between the lines. The launch of GPT-5.1 is a symptom of a deeper strategic shift. OpenAI is fragmenting its product because the “one AI fits all” model has failed. GPT-5 was so disappointing that the company had to enable it again GPT-4o as an option the next day.

In Xataka | OpenAI has never been more ambitious. And he’s never been so close to not being able to pay his debts.

Featured image | Xataka with Mockuuups Studio