OpenAI quietly released GPT-5.1 on Nov. 12, an update that abandons traditional performance data propaganda and focuses on users' emotional needs. Core upgrades include GPT-5.1 Instant (warmer and more talkative, supporting adaptive reasoning) and GPT-5.1 Thinking (optimizing thinking time allocation), providing eight chat style presets (Professional, Candid, and Quirky are added), and allowing fine-tuning of features such as warmth and brevity. Security Assessment adds Mental Health and Emotional Dependency dimensions, with some indicators slightly rolled back. Paid users can use it gradually, and it supports regression to the old model within 3 months, emphasizing the transformation of AI from a tool to a user-aware partner.
- This summary is generated by AI analyzing the content of the article for informational purposes only.
This morning, OpenAI officially released GPT-5.1, but the way it was released this time made people feel a hint of unusual. Looking through the official blog, I couldn't find a score comparison chart, no performance improvement XX% data, and even faster and stronger this kind of regular propaganda words are pitifully small.
This isn't quite the OpenAI we're familiar with, until we saw a quote from the official blog that said, as we've heard so clearly from our users, good AI should not only be smart, but also enjoyable to talk to.
From tool to partner, the emotional evolution of AI
The most striking thing about GPT-5.1 is not the technical parameters, but the fact that it finally looks like a real person. The new version provides eight style presets, allowing us to meticulously tune the AI's chatting style. Chatting with the AI is no longer a cold question and answer, but it has become a terse, understand the atmosphere, will pick up the conversation, and can even accompany you to make small talk in all parts of the world.
It seems that after the last GPT-5 word of mouth was cold, OpenAI finally listened to persuasion and publicly admitted for the first time: it is useless to just brush the list, what users want is an AI that can speak well, and practical and emotional value are indispensable.
Two core models, IQ and EQ double upgrade
The core of this update is the full upgrade of the two models GPT-5.1 Instant and GPT-5.1 Thinking.
GPT-5.1 Instant is the most frequently called model in ChatGPT. This time it has become warmer and more chatty. Based on early testing, it even shows some naughtiness from time to time, while keeping its answers clear and practical.
On a technical level, it now follows instructions more reliably and accurately answers the questions users really want to ask. Adaptive reasoning, introduced for the first time, allows it to intelligently decide to think first and give a more thorough and accurate answer when faced with a difficult question, and to remain extremely responsive when faced with a simple question.
As a more advanced reasoning model, GPT-5.1 Thinking also sees key optimizations. It now allocates thinking time more accurately, spending more time on complex questions with more thorough answers, and responding faster on simple questions with shorter wait times. What's more, its answers have become clearer and easier to understand, using less industry jargon and allowing us to understand complex concepts effortlessly.
In the words of Fidji Simo, CEO of OpenAI Apps, the core of this upgrade is a better combination of IQ and EQ. The model maintains a high IQ while dramatically improving emotional intelligence, allowing ChatGPT to have more natural conversations and empathy to meet users' individual needs in different scenarios.
Create an AI chat partner just for you
If model upgrades are the hard power, the leap in personalized experience is the soft power, which is another highlight of this update.
In addition to the original default, friendly and efficient, GPT-5.1 has added three new official styles:
- Professional: for formal situations such as work, writing, etc.
- Candid: more direct, no beating around the bush.
- Quirky: Getting funnier and more jumpy!
Even more surprisingly, GPT-5.1 allows users to fine-tune ChatGPT's features directly from the settings. You can precisely control how concise and enthusiastic the responses are, whether they are easy to navigate, and even how often emoji are used. ChatGPT may also proactively pick up on when you're channeling a particular tone in a conversation and ask whether to save that preference to a permanent setting.
In the past, custom commands have often been less than ideal, such as telling ChatGPT not to use a certain word and it using it anyway.GPT-5.1 has made significant improvements in this area, allowing custom commands to stick more consistently through multiple rounds of conversation.
Balance of safety and temperature
A warmer, more emotionally intelligent AI must also be safer. In the model introduction of GPT-5.1, OpenAI added two new humanization dimensions to security assessment for the first time:
Mental health assessment: how the AI responds to signs of isolation, paranoia, or mania that the user may exhibit.
Emotional attachment assessment: whether the AI's response promotes an unhealthy dependence or emotional attachment of the user to ChatGPT.
OpenAI candidly noted that GPT-5.1 Thinking regressed slightly compared to its predecessor in benchmark tests dealing with harassment, hate speech, and other content, and that the Instant model showed a slight regression in some assessments of emotional dependency. The company says it is working to improve these areas and has chosen to be transparent and open about this phenomenon, which is much more appreciated in a time of rapid AI iteration than a mere zero-failure campaign.
When will GPT-5.1 be available?
After updating to GPT-5.1, ChatGPT will switch to the latest model by default, no need to select it specifically.
- Paid subscribers (Pro, Plus, Go, Business): gradual push starting Nov. 12
- Free and non-logged-in users: will follow up with paid users when the push is complete
- Enterprise and Education Editions: have a 7-day Early Bird toggle switch (off by default), after which they will be upgraded uniformly
- API Developer: GPT-5.1 Instant and GPT-5.1 Thinking to go live on API later this week
To avoid a repeat of the embarrassment of users calling for the return of GPT-4 like the last time GPT-5 was released, OpenAI is offering a pill of regret this time: paid users can still choose to use the old GPT-5 model in the drop-down menu of the settings within 3 months.
From tools to partners for the future
GPT-5.1 is a major update that goes hand in hand with capability and experience.OpenAI is clearly telling us that the future of AI is not only about stronger parameters and higher scores, but also about a more understanding experience and more intimate interactions.
Fidji Simo has a poignant analogy in his article: if I could have complete control over my husband's idiosyncrasies, I'd probably get him to agree with me forever, but obviously, that's not a good idea.
The best AI should be like the best people in our lives, who listen, adapt, but also challenge us when necessary and help us grow. From a know-it-all jack-of-all-trades, to an exclusive partner that understands your chatting temperament and even helps you grow, this may be the future that GPT-5.1 really wants to unlock.
Comment List (8):
Load More Comments Loading...