ChatGPT Gets Intelligent UI: Answers You Can Actually Tap

On Tuesday, OpenAI started pushing its GPT-6 models to ChatGPT subscribers, along with a new feature called Intelligent UI that changes what an answer can look like. Instead of a wall of text, responses can now include graphics, tappable buttons, forms, charts and diagrams that readers use right inside the conversation. Paying subscribers on the Plus, Pro, Business and Enterprise plans get the upgrade first, running a model OpenAI calls Sol, while Free and Go users start receiving the rollout today on a model called Luna, according to the company's announcement. MacRumors reported the launch details on Wednesday, and gHacks covered how the rollout reaches every tier.

The Intelligent UI experience lives in the Chat tab, the regular conversation view. Ask for a recipe and the answer can include a cooking timeline with a widget that adjusts quantities when the guest count changes. Ask about a road trip and the stops land on a map. An explanation of a tricky idea can become a diagram with inputs to fiddle with, and a comparison can sit side by side instead of running down the page. For everyday tasks, ChatGPT can now build small tools on the spot: a retirement savings calculator, a dinner bill splitter, or a game played inside the chat, according to OpenAI's product post. The Intelligent UI feature works across web and mobile apps, so the interactive answer follows the conversation wherever it happens.

How the interface builds itself while the model writes

Behind Intelligent UI sits a library of native, streamable components. A compiler processes the interface while the model is still generating the answer, so buttons and charts appear progressively instead of popping in after the full response finishes. The same idea extends to reasoning: the model can begin answering while it keeps thinking through the rest, filling in the response in stages rather than holding everything back. OpenAI says wait time became part of the training, with the model learning to weigh how long someone sits idle before the first words land. The company is upfront that the model's design judgment still needs work, so some early Intelligent UI experiments may look rough around the edges.

OpenAI attaches figures to the speed claims, all from its own testing. The company says the Instant version of the new model starts answering questions that need web search 44 percent sooner, on average, than the previous generation's equivalent. On a batch of daily agent-style tasks, it began replying in the time the older model needed at a lower effort setting and still scored higher overall. Inside Intelligent UI, lookups got smarter too: the company reports the new model judges better when a web search is worth running and turns up backing for its claims more consistently. Each figure here comes from evaluations OpenAI ran itself.

OpenAI wants the answer to do the work

The move matters because of scale. OpenAI says more than a billion people use ChatGPT each week, a figure the company reports itself, which gives a product change like this a reach few standalone apps can match. RuntimeWire, which tracked the launch, framed the bet plainly: OpenAI is testing whether people will use answers that do things, instead of leaving the chat for a separate app. The timing lines up with a wider push across the industry. Microsoft is putting advanced AI onto personal computers with Nvidia-powered Surface devices and new safeguards for autonomous agents, while Apple is locking down the Mac against agent snooping. Interactive answers are OpenAI's bid to keep the conversation itself as the place where the work happens.

For now, Intelligent UI applies to the Chat experience only; the models behind ChatGPT Work and Codex stay as they are. Enterprise access depends on workplace administrator settings, and legacy desktop clients on Windows and macOS do not support the new capabilities, according to OpenAI's help page. Free and Go users should see the update land today, with Luna handling their conversations and Sol serving the paid tiers. What lands first is two conversation-tuned models with interactive answers attached, and the rest of the industry now has a very visible bar to clear.