GPT-6 and Intelligent UI in ChatGPT: What Changes for Staff on Free and Paid Plans

ChatGPT’s Intelligent UI started rolling out on October 7, 2026, and it changes what an answer looks like. Instead of text only, ChatGPT can now answer with diagrams, maps, comparisons, calculators and small interactive tools, and it picks the format on its own.

It came with a model change in the Chat tab. Paid plans moved to GPT-6 Sol on October 7. Free and Go moved to GPT-6 Luna the next day. This post covers what each plan got, what Intelligent UI does in practice, how to turn it down, what admins can control, and what agencies should tell nonprofit and association clients. Status is as of October 10, 2026.

The short version: everyone who uses ChatGPT’s Chat tab is on a GPT-6 model now, including staff on free personal accounts. Answers will look different, and some will include tools that run numbers for you. Those tools look finished and official, so staff should check the math before it goes into a budget or a board memo. Users can switch to a simpler style on the web. Enterprise admins can turn GPT-6 Sol access off. We found no separate admin switch for Intelligent UI itself.

What shipped on October 7 and 8

OpenAI’s announcement, GPT-6 and Intelligent UI for everyone, and its ChatGPT release notes give the same split. The release notes put it plainly: "Plus, Pro, Business, and Enterprise use GPT-6 Sol; Free and Go use GPT-6 Luna."

Intelligent UI works at every reasoning level from Instant through Extra High. The Pro reasoning option still runs on GPT-6 Astra and does not use it. The models behind Work and Codex did not change with this release. We covered the pricing logic behind Sol and Luna in GPT-6 Sol and Luna, when both were available only in Work, Codex and the API.

What Intelligent UI does in practice

OpenAI’s Intelligent UI help article says it "lets ChatGPT shape each answer with the right layout, visuals, and interactivity." Answers can include graphics, buttons, forms, charts and interactive tools. A comparison might come back side by side. A simple question still gets a text answer.

The examples in OpenAI’s announcement are everyday ones:

• A bicycle breakdown. A side-view diagram with tabs for frame, wheels, drivetrain, brakes and cockpit. Tap a part to see what it does.
• A Sunday roast. A recipe with cooking times, a guest-count control that recalculates the shopping list, and a timed checklist.
• A road trip. Stops on a map, with notes on detours worth taking.
• Small tools. A bill splitter, a retirement savings calculator and a retro game you can play in the chat.
• Learning topics. Explanations where you change an input and see the result, such as the Monty Hall problem.

OpenAI says it built "a library of native, streamable components" and a compiler that renders the interface while the model writes. So the layout appears piece by piece rather than all at once.

Some components keep their state. The help article says checklists can survive a refresh of the chat. It also says state does not carry across chat threads.

Answers start sooner

The second change is timing. OpenAI says GPT-6 "can interleave thinking with answering." It starts replying, keeps reasoning or searching, and adds findings without another prompt.

OpenAI gives two figures from its own testing. For questions that need web search, GPT-6 Instant starts answering 44% sooner on average than GPT-5.6 Instant. And GPT-6 at Extra High starts answering in about the same time as GPT-5.6 at Medium. OpenAI also says the model is better at deciding when to search. These are internal evaluations, and VKTR notes they have not been independently verified.

For staff, the practical effect is that the first lines of an answer may change as more arrives. Read to the end before copying.

How to turn Intelligent UI down

There is no full off switch that we could find. On the web, the help article gives one setting: Personalization, then Layout and Visuals, then Simple. OpenAI says this "reduces visual and interactive elements, though some may still appear."

Other ways to steer it:

• Custom instructions. OpenAI suggests lines such as “keep answers concise” or “use tables when comparing options.”
• Ask in the chat. You can request a plain text answer for one conversation. OpenAI says the exact layout can still vary.
• Pro reasoning. The Pro option runs on GPT-6 Astra and does not use Intelligent UI. OpenAI’s model help page lists that option for Pro, Business and Enterprise plans.

Intelligent UI is not available in the Work tab or in Voice. The help article does not list a Simple setting for the mobile apps. Plugins people have connected, such as Gmail, keep working.

Admin controls for Business, Enterprise and Edu

The admin picture is thin, and it differs by plan.

Enterprise. The Enterprise and Edu release notes say GPT-6 in Chat "uses the existing GPT-6 Sol access setting for Instant through Extra High." They add: "Admins can turn off access, and workspace and role permissions still apply." There is no new admin preview period this time.

Business. The Business release notes say GPT-6 is rolling out to eligible workspaces, depending on workspace access and supported clients. They do not name a control.

Edu. The October 7 entry is titled for Enterprise Chat and does not mention Edu. Edu admins should check their console rather than assume.

Free and Go. Personal accounts have no admin. Staff on free accounts got GPT-6 Luna and Intelligent UI on October 8 with nothing for IT to approve.

We found no admin setting that controls Intelligent UI on its own, separate from model access. If you run an Enterprise workspace, our ChatGPT Enterprise admin checklist covers the September settings that still apply.

Some OpenAI help pages have not caught up

As of October 10, several OpenAI pages still describe the old setup. The Free tier FAQ says Free users have GPT-5.6 Luna. The models page on learn.chatgpt.com says GPT-6 Sol and Luna "aren’t available in Chat." The release notes and announcement are newer and more specific, so we went with those.

This matters if your training materials or staff guides quote OpenAI’s help pages. Check the date on any page you copy from. Expect screenshots in internal guides to look out of date, too.

The October 5 speed change in Work and Codex

A separate change landed two days earlier. On October 5, Codex engineering lead Thibault Sottiaux posted on X that GPT-6 Astra and GPT-6.1 Sol are about 50% faster by default for subscribers. According to The New Stack, the change covers Codex and ChatGPT Work, plus partner tools that use Sign in with ChatGPT, such as OpenCode, Pi, Amp and Devin. PC Watch reports output rose from about 30 to about 50 tokens per second.

Both outlets describe it as day one of a 28-day run of daily improvements or usage resets. We did not find it in OpenAI’s changelog or release notes, so treat the figures as OpenAI staff statements reported by the press. It does not change the Chat tab. Dots, which we explained in what OpenAI dots are, run on GPT-6 Astra.

What changes for nonprofit and association staff

Most staff will notice the new look before anything else. Here is what deserves attention.

Calculators look official. A savings calculator or bill splitter built on the fly is still a model’s output. VKTR warns that such tools "can look authoritative." Check the numbers before they reach a grant budget, a dues model or a board deck.

Free accounts changed too. Staff using free personal accounts are now on GPT-6 Luna. The model is new, but the data rules for personal accounts did not change. Your policy on member data in personal accounts still applies.

Copying out may not work cleanly. OpenAI does not say how interactive elements copy, export or share. Test before staff build a habit around pasting them into documents.

Accessibility is not addressed. The help article says nothing about screen readers or keyboard use for interactive answers. If staff rely on assistive technology, test a few answers with them and use Simple if needed.

Model choice for API work is separate. None of this changes which model your integrations call. For that, see which OpenAI model to use.

How Anthropic and Google handle visual answers

ChatGPT is not first here. Both main competitors already show visuals inside answers.

Anthropic’s custom visuals help article says Claude builds diagrams, charts and interactive visuals in the conversation, and "You don’t need to turn anything on." The visuals are temporary by default. Claude’s artifacts are the persistent, shareable version. The article says custom visuals do not render in the iOS or Android apps.

Google launched visual layout and dynamic view with Gemini 3 in November 2025. Dynamic view has Gemini code a custom interface for the prompt. Google’s release notes introduced both as Labs experiments, and we found no later status update.

Claude’s version is also on by default for every plan, but it is labeled beta and skips the mobile apps. ChatGPT’s reaches web and mobile on every plan, and it arrived with a model change.

What agencies should tell clients

Keep the message short. Clients do not need a briefing on GPT-6. They need four points.

1. ChatGPT answers look different now, on free and paid accounts alike. That is expected.
2. Check any number from a built-in tool before it goes into a budget, proposal or report.
3. Use Simple under Personalization, then Layout and Visuals, if the visuals get in the way.
4. Enterprise admins can turn off GPT-6 Sol access. There is no separate Intelligent UI switch that we found.

Then update any staff guide with ChatGPT screenshots. If a client is weighing a paid plan, note that the visual answers are not a reason to upgrade. Free accounts got them too. Our OpenAI DevDay 2026 recap covers the plan changes that do differ by tier.

Frequently Asked Questions