ChatGPT can answer with charts, buttons and forms instead of plain text. GPT-6 reaches Plus, Pro, Business and Enterprise first, then Free and Go.
Until now a rep could assume ChatGPT would answer in prose, which meant the work of turning an answer into something usable, a table a manager will actually read or a grid you can scan before a call, stayed with the rep. Intelligent UI moves that step inside the answer. The catch is that the rep no longer chooses the shape: OpenAI says the format depends on the question, so the prompt is now the only control over whether an answer arrives as something you can act on or something you have to re-read. A chart also reads as more authoritative than a paragraph, which makes checking the numbers behind it a new habit rather than an optional one.
Two things shipped together and they are easy to confuse. GPT-6 is the model, the part doing the thinking. Intelligent UI is the delivery, the part deciding whether the answer arrives as a paragraph, a chart or a set of buttons you can tap. Sol and Luna are simply the two sizes of the model: Sol on the paid tiers, Luna on Free and Go. "In the Chat tab" matters because that is not the same surface as ChatGPT Work, which OpenAI says is unchanged. The short version: ChatGPT got quicker at searching and can now reply with a small interface instead of text, on the paid tiers first, and nothing about your Work setup moved.
OpenAI published "GPT-6 and Intelligent UI for everyone" on 7 October 2026, putting a new GPT-6 model behind the ChatGPT Chat tab and letting ChatGPT answer with an interface rather than a block of text. OpenAI describes the audience as "more than 1.2 billion weekly users." Answers can come back as graphics, tappable buttons, forms, charts and small tools built during the conversation, and OpenAI says the format "depends on what you're asking," noting that a simple text answer is sometimes best.
Announced: 7 October 2026.
Availability: "starts rolling out globally to ChatGPT Plus, Pro, Business and Enterprise tiers today in the Chat tab," powered by GPT-6 Sol. "Starting tomorrow, the rollout expands to Free and Go tiers," powered by GPT-6 Luna. "Enterprise availability depends on workplace admin settings."
Price: Not stated.
Regions: Not stated beyond "globally."
One speed number appears in the release. For questions that need a web search, OpenAI says GPT-6 Instant "starts answering 44% sooner, on average, than GPT-5.6 Instant," and that the model "makes better decisions about when to look something up." OpenAI also says GPT-6 is more resistant to attempts to bypass its safety training and is better at communicating "the limits of its own capabilities."
What it doesn't do: It does not touch ChatGPT Work or Codex. OpenAI states that "the models powering Work and Codex are not changing as part of this release." The release states no price, no API availability, no usage limits and no region list beyond "globally." It does not hand the rep a switch for the output format, since OpenAI says the format depends on the question, and it does not reach an Enterprise user whose workplace admin has not enabled it.
SDR — your account research will start coming back as grids and charts. That is faster to scan and easier to over-trust, so name your sources in the prompt and check any figure you plan to put in an email against the page it came from. A chart with a wrong number in it still looks finished.
AE — stop asking for a summary and start asking for a shape. "Give me a stakeholder grid with a risk column" now produces a different artefact than "tell me about the stakeholders," and the first one survives being read ninety seconds before a call. Re-run the two or three prompts you rely on most and keep the wording that returns structure.
Manager — the same question can now return different formats to different reps, so screenshots stop being comparable in coaching. If you compare rep output, standardise the prompt wording across the team first, and judge the reasoning rather than the formatting.
RevOps — two narrow jobs. Confirm the Enterprise admin setting, because OpenAI says Enterprise availability depends on it and your reps may be waiting on you rather than on the rollout. Then leave anything built on ChatGPT Work alone: OpenAI states Work and Codex models are not changing, so no re-testing is owed there. Do re-test anything downstream that parses ChatGPT output as text, because the output format is no longer predictable.
The quickest way to see whether an interface beats a paragraph is to hand ChatGPT something messy and ask for a structured read instead of a story. Run a discovery call transcript through this prompt and look at the shape of what comes back.
Two minutes, once a week. What changed in AI, and what to run because of it.