What's trending, explained.
What's trending, explained.
OpenAI has started rolling out GPT-6 to every ChatGPT user, alongside a new feature called Intelligent UI that changes what an answer looks like. Instead of a plain block of text, a response can now include interactive charts, tappable buttons, forms, sliders, maps, and even small tools built on the spot — a retirement-savings calculator, a bill splitter, or a simple game, all inside the conversation.
The rollout is split by plan and by model. ChatGPT Plus, Pro, Business, and Enterprise users began receiving GPT-6 with Intelligent UI on October 7, powered by a model OpenAI calls Sol. Free and Go users started getting the same interface on October 8, powered by a lighter model called Luna. The company says the model still answers in plain text when an interactive layout would add little, and the feature works across web and mobile.
Under the hood, GPT-6 is two models under one name. Sol is the more capable flagship for paying subscribers; Luna is the efficient version hundreds of millions of free users will actually meet. Both share the GPT-6 architecture, but the two-tier structure means the experience — and the quality — can differ between free and paid accounts.
OpenAI also claims GPT-6 responds faster on search-style questions: the company says GPT-6 Instant begins answering web-search queries 44% sooner, on average, than GPT-5.6 Instant. That figure comes from OpenAI's own testing, not an independent benchmark.
This is the biggest change to ChatGPT's interface since the chatbot launched. For years, AI answers have been paragraphs of text that users copy into other apps to act on. Intelligent UI moves the action inside the conversation itself — the model can render a working mini-tool at the moment you ask, rather than describing how to build one. If it works well, ChatGPT becomes less a question-answer box and more a workspace that assembles itself around your request.
The two-model split matters for a different reason. Luna is the model the vast majority of users will experience, and the paid/free quality gap now arrives with a new interface on top. Google already tried generative interfaces with Gemini, and early assessments warned of two side effects: interactive answers can burn through tokens faster, and visual elements give the model new ways to present wrong information convincingly. An attractive chart is valuable only if the numbers behind it are right.
There is also a business story underneath. OpenAI said days earlier that it would test a visual ad format for Free and Go users in the United States later this month, showing labeled ads during image generation, and its pricing page notes the Go plan may include ads. Interactive answers create far more visual real estate — and far more surface for advertising — than plain text ever did.
The rollout is global and ongoing, though access is not instant for everyone: a rollout start does not mean every eligible account sees the feature at the same time, and Enterprise availability depends on each organization's own settings. The models behind OpenAI's Work and Codex products are unchanged.
Watch two things. First, the safety picture: OpenAI's own system card — the technical report on how the models performed in its safety testing — acknowledges statistically significant regressions on self-harm and extremism evaluations versus GPT-5.6. Second, the ad test: if labeled ads land inside chat later this month for free-tier users, the conversation about what a "free" AI assistant costs will get louder. The interface is new; the economics underneath are just catching up.