OpenAI started rolling out GPT-6 in ChatGPT on October 7, 2026. The headline feature is what the company calls Intelligent UI: instead of replying only in text, the model can compose an answer from visuals and interactive pieces such as charts, tappable buttons, forms and small built-in tools.

Who gets which model

  • GPT-6 Sol powers ChatGPT for Plus, Pro, Business and Enterprise users, starting October 7.
  • GPT-6 Luna powers the Free and Go tiers, starting October 8.
  • Enterprise availability depends on workspace admin settings.
  • The rollout covers the Chat tab only. OpenAI says the Work and Codex models are unchanged in this release.

What Intelligent UI does

OpenAI says the model decides the format based on the question, and still uses a plain text answer when that is the most useful. Examples include visual plans and comparisons, interactive explanations of complex topics, and on-demand tools like calculators or bill splitters created inside the conversation.

Under the hood, OpenAI describes a library of native, streamable components and a compiler that renders the interface progressively while the answer is being generated.

Faster answers and safety claims

GPT-6 can interleave thinking and answering, so a response can start before reasoning finishes. OpenAI says GPT-6 Instant starts answering web-search questions 44% sooner on average than GPT-5.6 Instant. The company also claims stronger resistance to multi-turn jailbreak attempts and fewer unnecessary refusals.

A follow-up tweak

On October 29, OpenAI adjusted GPT-6 in Chat for Free, Go and logged-out users to give more concise and less flattering replies. For Plus and Pro, the change applies only to the fallback model used after GPT-6 message limits are reached.

How we got here

When ChatGPT launched in late 2022, it was a text box that returned text. Since then, chat assistants have steadily added other ways to answer: images, voice conversations, file analysis, and side panels for editing documents and code. Rivals have moved in the same direction. Anthropic’s Claude has offered Artifacts, separate windows for code, documents and small apps, since 2024, and Google introduced what it called generative interfaces with Gemini 3 in late 2025, where the model lays out a visual answer instead of a wall of text.

Intelligent UI is OpenAI’s version of that idea, built directly into the main chat. The difference it claims is that the model picks the format by itself, answer by answer, rather than waiting for the user to ask for a chart or a tool.

What this means for everyday users

For questions with a lot of moving parts, such as comparing three phone plans, planning a week of meals or splitting a group bill, an interactive answer can be easier to use than a long list. Being able to change one value and watch the result update is something plain text can’t do.

There are trade-offs. A calculator generated on the spot is only as accurate as the formula behind it, and an error is harder to spot inside a tidy interface than in a sentence you can read. Interactive elements can also be harder to copy into an email or notes app, and less accessible for people using screen readers if they aren’t built carefully. For anything involving money or health, it is worth checking the underlying numbers the same way you would check a text answer.

What we think

Moving beyond plain text is the right direction for chat assistants, and OpenAI’s decision to let the model choose the format makes the feature feel natural rather than bolted on. The real test is not the launch demos but ordinary use: whether simple questions still get short, fast answers, and whether the interactive ones save people time. The October 29 change toward shorter, less flattering replies suggests OpenAI is already tuning the balance based on how people use it.

What to watch next

  • Whether Intelligent UI reaches the Work and Codex models, which this release left unchanged.
  • How developers react if OpenAI opens the component library to apps built on its platform.
  • Independent tests of the 44% faster-answer claim on real search questions.