ChatGPT's answers just stopped being plain text

OpenAI pushed a new model into ChatGPT this week, and for once the headline isn’t really the model. It’s what ChatGPT does with its answers now. GPT-6 arrived alongside a feature called Intelligent UI, which lets the chatbot build small interfaces instead of typing everything out as a paragraph: diagrams you can tap through, calculators you can type numbers into, maps with stops marked on them.

It’s the clearest sign yet that the big AI labs think the text box has run its course as the only way to talk to a chatbot. Google has been nudging in the same direction with Gemini 3 in AI Mode. This week it’s OpenAI’s turn, and the rollout touches every paying ChatGPT tier plus the free one.

What actually shipped

OpenAI is releasing GPT-6 in ChatGPT worldwide, introducing the new Intelligent UI that combines text, visuals, and interactive features, with the update rolling out beginning in the Chat tab for Plus, Pro, Business, and Enterprise users, extending to Free and Go users a day later. In Chat, GPT-6 runs on GPT-6 Sol for the plans in this rollout and on GPT-6 Luna for Free and Go.

The pitch is simple: let the model decide how to lay out its own answer. OpenAI says it trained GPT-6 to decide how each response is laid out based on the question, and responses can include buttons, forms, and charts, while ChatGPT can still answer in plain text when that’s the most useful format.

OpenAI’s own demos lean into that. One example shows ChatGPT producing a diagram of a seven-speed bicycle with interactive buttons that highlight individual components, and another shows the chatbot walking users through Mahjong with tile visuals sorted into categories you can scroll through. A product manager at the company framed the problem plainly: “ChatGPT has predominantly been a text-based interface,” but when it comes to helping users “get things done… the most helpful answers aren’t just text.”

You can also ask it to build something small on the spot. OpenAI’s help page gives examples such as a calculator to explore how savings could grow, a bill splitter for dinner with friends, or a game you can play inside the conversation. If you’d rather not have any of that, there’s an escape hatch: users who prefer fewer visuals will be able to adjust the feature’s output, the company said.

It’s not available everywhere at once. The Intelligent UI feature is gradually being introduced on the web and supported, updated apps, at thinking levels from Instant to Extra High, though the Pro thinking level continues to run on GPT-6 Astra and doesn’t support the feature, and neither do the older ChatGPT desktop apps for macOS and Windows. If you’re on an older desktop build, you’ll still get an answer, just not the diagram.

Why it starts answering before it’s “done”

The other change bundled into GPT-6 is speed of a specific kind. GPT-6 can begin answering while it continues to think, and Intelligent UI also appears progressively: a library of native components and a compiler display the interface as the model generates it. In practice that means the layout builds itself in front of you rather than popping in once the whole response is ready, which matters more for patience than for accuracy.

What decides whether you get a diagram or a paragraph is the question itself. The format depends on the question: a comparison might work best side by side, an explanation might be easier to understand through an interactive diagram, and when a simple text answer is most useful, ChatGPT can give you that.

How this stacks up against Gemini and Copilot

OpenAI isn’t first here, it’s catching up to something Google already shipped. When Google rolled out Gemini 3 inside AI Mode on Search, one standout feature was how Gemini 3 can display “immersive visual layouts and interactive tools and simulations, all generated completely on the fly based on your query.” Early hands-on coverage described asking Gemini to explain a physics concept and getting back a working simulation rather than a wall of text, the same basic idea OpenAI is now building into ChatGPT.

Microsoft’s Copilot hasn’t gone this far yet. Independent comparisons this year still describe it as leaning on DALL-E 3 integration for image generation rather than building interactive interfaces on the fly, and its multimodal handling is generally rated behind Gemini’s for video and documents. For now, interactive, generate-as-you-go answers are a two-horse race between OpenAI and Google, with Microsoft still playing a more traditional assistant role inside Office.

Assistant	Interactive/visual answers	Entry price	Context window
ChatGPT (GPT-6, Intelligent UI)	Diagrams, calculators, forms, generated per question	Free, Go €7.99/mo, Plus €22.99/mo	Varies by tier
Gemini (AI Mode, Gemini 3)	Immersive visual layouts, live simulations	Free, Google AI Pro ~€21.99/mo	Up to 1 million tokens on paid tiers
Microsoft Copilot	DALL-E 3 images, limited interactivity	Free, Copilot Pro varies by market	128,000 tokens

Prices above are listed in EUR where the provider publishes an EU figure; OpenAI lists ChatGPT Plus at EUR 22.99 and Go at EUR 7.99 across euro storefronts, while Google AI Pro is listed at €21.99 in France, according to pricing trackers checked this autumn. Copilot’s consumer pricing varies more by bundle and market, so treat that line as indicative rather than official.

Entry-level monthly price of major AI assistants (EUR)

What it costs to get GPT-6

GPT-6 itself isn’t a separate purchase, it’s the model now sitting behind the plans that already existed, though which exact version you get depends on what you pay. Go at $8/month defaults to GPT-6 Luna for unlimited everyday text chats without Sol access, while Plus at $20/month removes ads and moves you up to GPT-6 Sol reasoning with broader access to deep research, Codex and projects. Pro sits well above both: Pro costs $100 or $200 a month depending on usage tier, aimed at people who hit Plus limits regularly rather than anyone wanting Intelligent UI specifically, since the feature itself is available across tiers, not gated to Pro.

Who should actually care

If you use ChatGPT mainly for quick text answers, emails, summaries, this changes very little day to day. Where it matters is the stuff that’s always been awkward to explain in prose: working out loan repayments, planning a trip with several stops, understanding how a mechanical part fits together. Those are the cases where a diagram or a small calculator genuinely beats three paragraphs of explanation.

It’s also a signal about where the big three assistants are heading. Chat interfaces that only produce text are starting to look dated next to ones that can sketch, calculate and simulate on request. Anyone picking between ChatGPT, Gemini and Copilot for everyday use should weigh this alongside price and ecosystem fit, not just raw model benchmarks.

Also this week

Google gave Gemini a new job title: coworker. Google’s new Gemini agent gives each coworker agent its own Workspace account, email address, calendar and entry in the company directory, acting under its own identity rather than an employee’s. It’s currently available only to enterprise customers in private preview, with no word on a wider consumer release, so this is one to watch rather than try yet, but it signals where workplace AI assistants are headed next: less chatbot, more standing team member with its own inbox.

Apple is set to hold a dedicated smart home event on October 13. Bloomberg’s Mark Gurman has said Apple plans to hold an “online home event” to introduce an all-new smart home hub, a new HomePod mini, and a new Apple TV 4K, all built around Siri AI. Earlier Bloomberg reporting cited by tech publications puts the anticipated price point for the hub around $350, though Apple hasn’t confirmed pricing or even the products themselves yet. If the leaks hold, it’s Apple’s first real push into the smart display category that Amazon and Google have occupied for years.

What this means for you

If you’re already paying for ChatGPT Plus or Pro, Intelligent UI shows up automatically, no action needed. Try it on something you’d normally Google, like “explain how a heat pump works” or “help me plan a three-stop road trip”, to see what it actually changes.
If you’re on an older ChatGPT desktop app for macOS or Windows, update it. The feature is not supported on the legacy desktop clients.
If you’re choosing between ChatGPT Plus and Google AI Pro for everyday visual/interactive answers, know that Google got there first through AI Mode in Search; it’s worth a side-by-side test of your actual use cases before paying for either.
Hold off on any Apple smart home purchases until October 13. If the leaked $350 price for the hub is close, it will land well above a HomePod mini and well below a MacBook, so it’s worth seeing the real specs before deciding.

ChatGPT’s answers just stopped being plain text