Follow ZDNET: Add us as a preferred source on Google.
ZDNET’s key takeaways
- OpenAI’s GPT-5.1 includes two brand-new models.
- Users can also choose from new personalities.
- The updates are meant to make ChatGPT more personalized.
OpenAI updated ChatGPT with its highly-anticipated GPT-5 model over the summer, which was supposed to streamline the user experience by making the chatbot automatically pick the right model based on the input query. However, people immediately missed the ability to personalize the experience to their needs — and this GPT 5.1 release promises to bring that choice back.
On Wednesday, OpenAI upgraded its GPT-5 series of models in ChatGPT to include two brand-new models: GPT-5.1 Instant and GPT-5.1 Thinking. These models aim to make using ChatGPT a more seamless experience, featuring what OpenAI said are “smarter” and “warmer” conversations.
Also: OpenAI says it’s working toward catastrophe or utopia – just not sure which
This release was also accompanied by new customization tools and additional personalities for the chatbot. For the full roundup of new features, keep reading below.
(Disclosure: Ziff Davis, ZDNET’s parent company, filed an April 2025 lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.)
The new models
GPT-5.1 Instant
When interacting with ChatGPT daily, most of your prompts will likely be relatively simple in nature and, therefore, not require extensive reasoning. As a result, GPT-5.1 is OpenAI’s most widely used model, having undergone significant upgrades.
Also: How to turn off Gemini in your Gmail, Photos, Chrome, and more – it’s easy to opt out of AI
For starters, it has now been upgraded to be “warmer” by default, and more conversational. OpenAI said in the blog post that early testers were surprised by how playful the model can be while remaining useful. GPT-5.1 is also better at instruction-following, which means it can better understand what you were requesting with your prompt, and, as a result, provide a more helpful answer that actually addresses your prompt.
Lastly, for the first time, the model can employ adaptive reasoning, which enables it to think before responding to more challenging prompts, theoretically resulting in more accurate responses for complex tasks while staying speedy for easier ones. As a result, OpenAI reported that it performed significantly better on the AIME 2025 (an industry-standard mathematical benchmark) and Codeforces (a coding benchmark).
ZDNET has yet to test GPT-5.1 but will update this story once we do.
GPT-5.1 Thinking
GPT-5 Thinking was built to tackle more complex tasks and therefore spend more time reasoning or thinking through a prompt ot generate an answer. To optimize the experience, GPT-5.1 can better adapt the amount of thinking to the complexity of the question, balancing speed and computational power with helpful responses.
Also: Is ChatGPT Plus still worth $20? How it compares to the Free and Pro plans
An internal benchmark posted by OpenAI shows that GPT-5.1 better adjusts for thinking time than its predecessor, thinking nearly twice as fast on the fastest tasks and twice as slow on the slower tasks. The company said the responses were also clearer and contained less jargon, making it OpenAI’s most capable model yet.
How to try GPT-5.1 models
GPT-5.1 Instant and Thinking are both rolling out to users now, with priority going to paid subscribers, including Pro, Plus, Go, and Business plans, followed by free and logged-out users. The company said the rollout will be phased over the next few days to maintain stable performance for everyone. OpenAI plans to update GPT-5 Pro to GPT-5.1 soon.
GPT-5 Auto will remain available in ChatGPT and will continue to automatically select the best model for the chatbot based on the prompt. The option to select from the new models is intended for specific use cases, or if you prefer a smoother and more natural tone.
Also: How to use ChatGPT: A beginner’s guide to the most popular AI chatbot
On Thursday, Enterprise and Edu plans will get a seven-day early-access toggle (off by default), but then GPT-5.1 will become the default model. Both GPT-5.1 Instant and Thinking will be released in the API later this week.
The legacy GPT-5 (Instant, Thinking, and Pro) will be available in ChatGPT’s legacy models dropdown for paid subscribers for three months, allowing users to compare the two model families. This sunset period won’t impact other legacy models.
New personalization options
In addition to trying the new models, you can also personalize ChatGPT’s tone and style with more precision than ever.
Also: I’ve tried a lot of AI smart glasses and the Even G2 are the only ones I’d wear all day
For starters, you can change ChatGPT’s tones and personality to more options: Default, Friendly (formerly Listener), and Efficient (formerly Robot) remain (with updates), and users now have Professional, Candid, and Quirky as well. OpenAI stated that these options are designed to better reflect diverse user preferences.
The new personalities apply to all the new models; the formerly introduced Cynical and Nerdy personalities will continue to be available in the dropdown personalization settings. These settings are rolling out today.
Beyond personality, you can now also tune ChatGPT’s characteristics. Now, in the personalization settings, you can adjust how concise, warm, or scannable the chatbot’s responses are. This also tackles one of people’s biggest pet peeves: how often ChatGPT uses emojis. The best part is that it can offer to update your preferences based on what you indicate in conversation, without you having to manually navigate to settings. This will be available later this week as an experiment for a limited number of users.
Also: A new Chinese AI model claims to outperform GPT-5 and Sonnet 4.5 – and it’s free
Lastly, GPT-5.1 is better at understanding custom instructions for more personalized results, which can also be used to tweak ChatGPT’s tone and outputs. OpenAI said updates you make apply to all of your chats, including ongoing chats — a change from the prior system in which these adjustments only applied to conversations thereafter.
Comments are closed.