GPT-6 comes to free ChatGPT users, with answers you can tap and explore

If you open ChatGPT on a free account today, there is a good chance it will look different. On October 7, OpenAI started rolling out a new GPT-6 model inside ChatGPT, together with a feature it calls Intelligent UI. Paying customers got it first. Starting Thursday, October 8, the rollout expands to the Free and Go tiers. That means the next generation of OpenAI's models is no longer something only subscribers and developers talk about. It is landing in front of the more than 1.2 billion people who, according to OpenAI, use ChatGPT every week.

The headline is not only a smarter model. It is a different kind of answer. Instead of a wall of text, ChatGPT can now reply with charts you can play with, buttons you can tap, forms you can fill in, maps, and small tools it builds on the spot. OpenAI says it "marks a significant shift in how people interact with ChatGPT," and after reading the announcement and the first reports, I think that is not just marketing.

The three things you need to know

• Intelligent UI. GPT-6 in ChatGPT can compose an answer out of text, visuals and interactive elements. Depending on the question, a reply can include graphics, tappable buttons, forms, charts or a complete little tool, such as a savings calculator or a bill splitter, right inside the conversation.
• Faster answers. ChatGPT can now start answering while it is still thinking. OpenAI says that for questions that need a web search, GPT-6 Instant starts answering 44 percent sooner on average than GPT-5.6 Instant.
• Who gets it. Plus, Pro, Business and Enterprise users started getting GPT-6 Sol on October 7. Free and Go users get GPT-6 Luna starting October 8. If the new look is too much, you can turn the visuals down in the settings.

What Intelligent UI actually is

Until now, ChatGPT has mostly been a text box with a text answer. You could get images, tables and code blocks, but the basic format was always the same: you ask, it writes. Aarush Selvan, a product manager at OpenAI, put it plainly in a call with journalists reported by TechCrunch: ChatGPT "has predominantly been a text-based interface," but for everyday things like planning a recipe, a trip or learning something new, "the most helpful answers aren't just text."

With Intelligent UI, GPT-6 decides for each question what the best format is. OpenAI says the model was trained to compose responses from text, visuals and interactive elements and to choose how they fit together. A comparison might be shown side by side. An explanation might become an interactive diagram. And when a short text answer is the most useful reply, ChatGPT is supposed to simply give you that.

The examples OpenAI and the first reports show are very practical. Ask how a seven speed bicycle is designed, and ChatGPT draws a diagram of the bike with buttons for the frame, wheels, drivetrain, brakes and cockpit, and each button explains its part. In the TechCrunch demo, Selvan asked ChatGPT to explore how an airplane wing generates lift, and the chatbot produced a set of diagrams. The Verge describes an answer that teaches you Mahjong by drawing every tile and sorting them into categories you can scroll through.

Planning works the same way. OpenAI shows a Sunday roast plan where the cooking timeline sits next to the recipe, and a shopping list that recalculates the amounts when you change the number of guests. A road trip can appear on a map with the stops marked. MacRumors also mentions checklists, integrated maps and multiple choice follow up questions that help narrow down what you actually want.

Tools built in the moment

The part I find most interesting is the third use case. You can ask ChatGPT to build a tool for the task in front of you. OpenAI's examples are a calculator that shows how savings could grow, a bill splitter for dinner with friends, and a small game you can play inside the chat. The Verge adds a retirement savings calculator and a retro game to the list.

That sounds small, but it changes what a chatbot is. Until now, if you needed a mortgage calculator, a packing list or a quiz, you went looking for an app or a website, or you asked an AI to write code that you then had to run somewhere. Now the tool simply appears in the answer, ready to use. OpenAI says this out loud at the end of its announcement: for decades, people had to learn how to use software and navigate fixed interfaces. "Instead of people adapting to software, software will adapt to people."

How it works under the hood

OpenAI also explains a bit of the technology. The company built a library of native, streamable interface components and a compiler that processes the interface while the model is generating it. The library gives every answer a familiar design, so ChatGPT still looks like ChatGPT on the web and on mobile. The model decides how to combine the pieces. Because of the compiler, the interface appears progressively as the answer is written, instead of after the whole response is finished.

To teach GPT-6 good design judgment, OpenAI says it expanded its training methods and evaluated the interfaces the model creates for clarity, usefulness and completeness. The model learned when interactivity helps and when plain text is enough. OpenAI is also honest that this is not finished: there is "still work ahead to improve the model's design judgment and expand what it can create."

Answers that start while the model is still thinking

The second big change is speed. Reasoning models can solve harder problems, but until now you had to wait until the model finished thinking before you saw anything. GPT-6 in ChatGPT can interleave thinking and answering. OpenAI says it trained the model to take the user's waiting time into account and to build the answer from several partial responses, each adding useful information, while keeping the full answer as coherent as if it had been written in one go.

OpenAI gives two numbers from its own internal tests. For questions that need a web search, GPT-6 Instant starts answering 44 percent sooner on average than GPT-5.6 Instant. And in an internal evaluation of high value everyday agentic tasks, GPT-6 Extra High begins answering in the same time as GPT-5.6 Medium, while scoring better overall than GPT-5.6 Extra High. OpenAI also says GPT-6 makes better decisions about when to look something up and finds supporting information more reliably. These are company figures, not independent benchmarks, so I would treat them as a direction rather than a final verdict.

Safety claims

OpenAI says GPT-6 builds on several safety advances from GPT-6 Astra, the first GPT-6 model it introduced for paying customers last month. Compared with GPT-5.6 Sol, the company says GPT-6 resists attempts to bypass its safety training better, especially attacks that adapt across several turns of a conversation. It was trained to use the conversation history to recognize risks that may not be visible in a single prompt, and to avoid refusing harmless requests. OpenAI also says the model is better at saying when it lacks the information or tools to answer. The details are in the system card that OpenAI published alongside the launch.

Who gets what, and when

Here is the rollout as OpenAI describes it:

• Plus, Pro, Business and Enterprise: GPT-6 with Intelligent UI started rolling out globally on October 7 in the Chat tab, powered by GPT-6 Sol. For Enterprise, availability depends on the admin settings of the workspace.
• Free and Go: the rollout starts on October 8, powered by GPT-6 Luna. The Verge describes Sol as the mid range model and Luna as the more efficient one.
• Work and Codex: nothing changes there. The models behind Work and Codex are not part of this release.

According to OpenAI's help center, the new capabilities run on the web and in updated mobile apps, from Instant up to Extra High. They are not available with Pro effort, which stays powered by GPT-6 Astra. The older ChatGPT desktop apps for macOS and Windows do not support Intelligent UI, so OpenAI recommends the web version there. There is no separate usage quota for Intelligent UI: the normal model, tool and plan limits still apply.

If you prefer the classic look, there is a switch. On the web, open Settings and turn off "Layout and visuals" to reduce visual elements. OpenAI notes that some visuals may still appear and that the setting does not turn ChatGPT into a pure text mode. Changing the look of one conversation does not change your other chats.

Why it matters

This launch matters for two reasons. The first is reach. Most people never pay for an AI subscription. When a new model generation lands in the free tier, it becomes the default experience for hundreds of millions of people overnight, from students doing homework to people planning dinner. Whatever ChatGPT does by default quickly becomes what people expect from every AI assistant.

The second reason is the format itself. A chatbot that answers with interactive tools competes with a lot of simple software and websites: calculators, recipe sites, trip planners, flash card apps and many small utilities. If the answer is the app, fewer people need to click through to anything else. That is great for users who want a quick result, and a real question for publishers and app makers who live from that traffic. Google has been experimenting with generated interfaces in Search as well, so this is clearly the direction the big players are moving in.

There is also an open question about trust. A nice chart looks convincing, even when the numbers behind it are wrong. With plain text, many people have learned to double check. With a polished interactive widget, it is easier to forget that the content still comes from a language model. OpenAI says GPT-6 is better at admitting what it does not know, but checking the important numbers stays your job.

Dany's take

I like this more than I expected. For learning, visual answers are simply better. When I want to understand how something works, an interactive diagram beats five paragraphs of text every time, and a calculator that appears exactly when I need it is genuinely useful. The faster first answer is also a quality of life upgrade that people will notice without reading any benchmark. What makes me a bit careful is the bigger picture. Every widget that appears inside ChatGPT is one less reason to visit a website or open another app, and that shifts even more of the internet into one chat window owned by one company. I also want to see how well the design judgment works in daily use: a chart for a simple yes or no question would get annoying fast, and I am glad there is a setting to dial it down. My advice: try it today with something you actually want to learn or plan, check the numbers in anything important, and decide for yourself whether you like the new look. Would you use ChatGPT more with visual answers, or do you prefer plain text? Tell me on Reddit, X or YouTube.

Source

Primary source: OpenAI, "GPT-6 and Intelligent UI for everyone". Availability and settings: OpenAI Help Center, "GPT-6 and other models in ChatGPT". Coverage: TechCrunch, The Verge and MacRumors.