Frontier models

GPT-6 and Intelligent UI in ChatGPT

On 7 October 2026, OpenAI started moving the main chat in ChatGPT to GPT-6 and added Intelligent UI (OpenAI, GPT-6 and Intelligent UI for everyone). Intelligent UI lets an answer include pictures, maps, checklists, and controls you can tap or change, instead of text alone. No special prompt is needed, and plain text answers still happen when they fit (OpenAI Help, GPT-6 and other models in ChatGPT). OpenAI says more than 1.2 billion people use ChatGPT each week.

Status: Feature, a ChatGPT feature, not an API model
Released: 7 Oct 2026
Last checked: 12 Oct 2026newest source date on this page

What it looks like

OpenAI's launch post shows the range. Ask about the design of a 7-speed bike and you get a side-view diagram of its five connected systems. Ask for help with a Sunday roast and the plan comes with a guest count you can change, which recalculates the shopping quantities, plus a cooking checklist. Other examples include road-trip stops on a map, help building a wardrobe, and visual explainers for the central limit theorem, the Monty Hall problem, GDP, and drone photography.

You can also ask it to build a small tool inside the chat. OpenAI's examples are a bill splitter for dinner with friends, a retirement savings calculator, and a retro game you can play in the conversation.

How it works

In OpenAI's words: "We built a library of native, streamable components, along with a compiler that processes the interface as the model generates it". A component is a ready-made building block for the screen, such as a chart, a map, or a button. Streamable means each piece can appear while the answer is still being written. The compiler is the software that turns what the model writes into the interface you see. OpenAI says this lets the interface appear "progressively as the model generates it, without waiting for the entire response to be complete".

OpenAI trained GPT-6 to choose the content, layout, visuals, and interaction for each answer. It graded the interfaces the model built for "clarity, usefulness, and completeness". OpenAI also says there is "still work ahead to improve the model's design judgment".

GPT-6 can also start answering while it keeps thinking, then add detail as it goes. OpenAI says GPT-6 Instant starts answering questions that need a web search 44% sooner on average than GPT-5.6 Instant. That is OpenAI's own measure, not an outside one.

Where it works and where it does not

	Works	Does not work
ChatGPT area	Chat tab	Work and Codex (their models are not changing)
Effort setting	Instant, Medium, High, Extra High	Pro
Apps	Web, updated mobile apps	Older Mac and Windows desktop apps
Developers	-	No API. OpenAI announced nothing for developers

The effort setting is the slider that sets how long ChatGPT thinks before it answers. Pro effort runs GPT-6 Pro, which is powered by GPT-6 Astra, and Intelligent UI is not available there. OpenAI says the update "applies to the Chat experience in ChatGPT".

Intelligent UI has no separate usage limit. Your plan's normal limits apply. On the web, Settings has a "Layout and visuals" switch. Turning it off cuts back the visual elements, but some may still appear, and it does not switch ChatGPT to a text-only mode.

Who gets it, and when

Plus, Pro, Business, and Enterprise started getting it on 7 October. For Enterprise, it depends on admin settings. OpenAI's launch post says Free and Go would follow on 8 October, but OpenAI's help page (checked 12 October) still says Free and Go get GPT-5.6 Luna, not GPT-6.

On Plus, the Medium and High effort levels are included. Extra High and Pro are not.

Which model is behind it

OpenAI's launch post says paid plans get GPT-6 Sol and Free and Go get GPT-6 Luna. In OpenAI's words: "Both are tuned for everyday conversation".

OpenAI's own help page disagrees on one point. As of 12 October it still says Free and Go users get GPT-5.6 Luna, both for normal answers and for Think. The launch post, OpenAI's developer forum post, The Verge, and MacRumors all say GPT-6 Luna (OpenAI Developer Community, The Verge, MacRumors). The two OpenAI pages still disagreed when this page was written.

How the models score

Artificial Analysis tests the API versions of these models, not the ChatGPT versions. OpenAI says the ChatGPT models are tuned for everyday conversation, so in-app results may differ from the scores below.

Artificial Analysis, max effort, fetched 8 Oct 2026	GPT-6 Sol	GPT-5.6 Sol
Intelligence Index	48	47
Cost per Index task	$1.04	$1.99
List price, input / output per million tokens	$2 / $10	$4 / $20

The Intelligence Index is one score Artificial Analysis builds by running the same set of tests on every model. GPT-6 Sol scores about the same as GPT-5.6 Sol at about half the cost per task (Artificial Analysis, GPT-6 Sol, Artificial Analysis, GPT-5.6 Sol). Artificial Analysis now marks GPT-6 Sol as deprecated on its site. OpenAI has since released GPT-6.1 Sol for the API.

GPT-6 Luna scores 38 on the same index, fetched 8 Oct 2026 (Artificial Analysis, GPT-6 Luna).