OpenAI releases GPT-5.1 with personality presets

Background

A chatbot's reply has two parts: what it says and how it says it. For a general assistant the second part matters more than it first appears — a flat, formal register suits some tasks and grates in others, and users rarely want to rewrite the same instruction about tone every time they start a conversation. Separately, AI labs have been splitting flagship models into a fast everyday model and a reasoning model that spends extra computation on a problem before answering. The GPT-5.1 release did both things at once ^[1].

What happened

On 12 November 2025 OpenAI released GPT-5.1 Instant and GPT-5.1 Thinking in ChatGPT ^[1]. The update made the assistant's default tone warmer and added eight personality options ^[1][2]. As the names suggest, the two variants have different jobs: GPT-5.1 Instant is the everyday model, while GPT-5.1 Thinking is a reasoning model built for more complex tasks ^[1]. Paid users received the new models first, with free users following ^[1].

Key details

OpenAI described GPT-5.1 Instant as its most-used model, now warmer and better at following instructions, and GPT-5.1 Thinking as its advanced reasoning model — faster on simple tasks and more persistent on complex ones ^[3].

The personality presets were reworked rather than simply expanded. Default, Friendly and Efficient remained, and were joined by new Professional, Candid and Quirky options; Cynical and Nerdy were left unchanged ^[3]. That makes eight settings in total, each a shorthand for a way of speaking that the user picks instead of describing in a prompt.

The personalisation changes also took effect across all chats immediately, including conversations already under way, rather than applying only to new chats ^[3]. In practice a chosen tone behaves like an account-level preference rather than something attached to a single thread: switch it, and the conversation you are already in switches with it.

Why it matters

Two design decisions stand out. The first is that voice became a setting. Rather than asking users to coax a particular register out of the model with wording, OpenAI exposed tone as a small set of named options sitting alongside the model itself — closer to a preference panel than to a prompt-engineering technique.

The second is retroactivity. Most changes to a chat product arrive with the next conversation; here the new tone and personality choices reached existing chats at once ^[3]. For anyone using ChatGPT for ongoing work, that means the assistant can change character mid-thread, which is convenient when a task shifts from drafting to reviewing and awkward when consistency was the point.

The split between Instant and Thinking reflects the same balancing act. Simple requests are answered quickly, while harder ones get a model that keeps working at them ^[3], so the choice of variant becomes part of the user's decision about what kind of help they want.

Afterwards

A week later OpenAI added GPT-5.1 Pro and the GPT-5.1-Codex-Max coding model ^[1]. Within a short span the line-up covered everyday chat, deliberate reasoning, a higher-capability tier and a dedicated coding model, with personality presets cutting across it as a layer of their own.