OpenAI has launched Intelligent UI in ChatGPT, giving GPT-6 the ability to decide whether a response should appear as text, graphics, clickable buttons, forms, charts, or directly usable interactive tools, and to combine those formats inside a single answer.
In its official announcement, OpenAI said ChatGPT now has more than 1.2 billion weekly users. The rollout started on Oct. 7 U.S. time for paid plans including Plus, Pro, Business and Enterprise. Free and Go plans followed on Oct. 8.
Responses are no longer limited to blocks of text
For the past several years, ChatGPT answers have largely looked the same: a wall of text that users had to read and interpret on their own. Intelligent UI changes that setup. OpenAI said GPT-6 has been trained to determine which presentation format best fits a given query.
The company said the format depends on the question itself. Comparison-heavy prompts may be clearer in a side-by-side layout. Explanations of how something works may fit an interactive visual. If plain text is still the most useful option, ChatGPT will continue to answer in text.
OpenAI’s examples were practical ones: a Sunday barbecue plan displayed next to recipes, road-trip stops placed directly on a map with notes on where a detour may be worthwhile, a savings growth calculator built on the spot, a bill-splitting tool for group dinners, and even a game that can be played inside the conversation.
The company’s mockups also showed fall outfit suggestions paired with photos and color swatches.
The interface is generated while the answer is being produced
OpenAI said the system relies on engineering work underneath the surface. The team built a native component library that can stream output in real time — a set of prebuilt buttons, charts and form elements that gives each response a consistent design base, while the model decides how those pieces should be arranged.
The company also said it built a compiler that processes the interface into a visible screen while the model is still generating content. Because the output is streamed, content appears progressively on screen instead of waiting for the full response to finish first.
GPT-6 changes how answers are delivered
OpenAI said earlier reasoning models could solve harder problems, but users often had to wait until the model finished thinking before seeing any answer. With GPT-6, thinking and answering are interleaved. The model can provide an initial response, continue reasoning, and then add more detail. OpenAI said user wait time was also factored into training.
OpenAI says the feature is still incomplete
In the announcement, OpenAI acknowledged that the model still needs better judgment in interface design and that the range of content it can generate needs to expand. Users who prefer fewer visual elements can reduce them through settings, similar to other ChatGPT personality controls. Plain-text responses have not been removed.
Free and paid users are not getting the same model. OpenAI said it first released the initial batch of GPT-6 models to paid customers last month, and this update expands the reach. Paid plans are powered by GPT-6 Sol. Free and Go plans had used GPT-5.6 Luna as the default model since August, and are now switching to GPT-6 Luna.
OpenAI closed the announcement with a broader product statement: in the past, people had to learn how to use software; in the future, software will adapt to people.