On October 7, OpenAI began rolling out GPT-6 with Intelligent UI in ChatGPT, allowing responses to combine text with charts, forms, buttons, and interactive tools. Instead of merely reading an explanation and typing a follow-up question, users can modify data or explore parts of the content directly within the conversation.
This is an update to the Chat experience, not a wholesale overhaul of models across Work and Codex. The highlight of the announcement lies in how ChatGPT presents and enables the use of responses, alongside improvements to the time it takes to begin responding.
Responses can become ready-to-use tools
According to OpenAI, GPT-6 is trained to choose how to combine text, visuals, and interactive components based on the prompt. A comparison request can be presented side-by-side; a complex concept can be explained using a diagram that allows selecting individual parts. If plain text is the most appropriate approach, ChatGPT can still respond with standard text.
In a weekend meal example, ChatGPT generated a control to adjust the guest count and update ingredient quantities accordingly. With a seven-speed bicycle, users can select the frame, wheels, drivetrain, or brakes to see each system’s role. OpenAI also highlighted use cases such as bill-splitting tools, savings growth calculators, and games directly inside the chat.
This change allows users to test multiple scenarios within the same response. For instance, rather than re-entering a prompt every time the attendee count changes, they can adjust a value directly. However, not every prompt generates a chart or form, and users do not need a special prompt to activate the feature.
Technically, OpenAI built a UI component library along with a compiler that renders the interface while the model generates content. As a result, parts of the response can appear progressively, rather than waiting for the entire interface to finish generating. The company stated that training included evaluations of the interface’s clarity, usefulness, and completeness; its ability to choose designs still has room for improvement.
Starting to respond sooner does not mean finishing faster
GPT-6 can also begin responding while continuing to reason. The model delivers information in chunks and then supplements the results during processing, instead of keeping users waiting until all reasoning steps are completed.
According to OpenAI’s internal evaluations, for prompts requiring web searches, GPT-6 Instant begins responding 44% faster on average compared to GPT-5.6 Instant. This figure was published by the company and has not been independently verified. It measures the time until the response starts appearing, and does not prove that the complete response or interactive charts finish 44% faster.
OpenAI also stated that the model has improved its ability to decide when to search and locate supporting information for answers. These findings are likewise based on the company’s own evaluations and should be distinguished from real-world performance across different types of prompts.
Rollout scope still depends on plans and devices
The October 7 announcement stated that the update begins with Plus, Pro, Business, and Enterprise, before expanding to Free and Go starting October 8. OpenAI described GPT-6 Sol as powering the aforementioned paid plans, while GPT-6 Luna serves Free and Go.
However, the Help Center currently still describes Free and Go as using GPT-5.6 Luna, which differs from the launch announcement. Therefore, the announced expansion schedule is not enough to confirm that all free accounts have received GPT-6. Enterprise users also depend on access permissions configured by administrators.
According to the Help Center, Intelligent UI works on the web and supported updated apps at reasoning levels ranging from Instant to Extra High, depending on plan entitlements. The feature is unavailable at the Pro level using GPT-6 Astra. Legacy ChatGPT desktop apps for macOS and Windows also do not support this experience.
Intelligent UI does not have a separate usage limit, but existing model, tool, and plan caps still apply. Web users can turn off the “Layout and visuals” toggle in settings to reduce visual elements; doing so does not guarantee that every response will revert to text only.
TechRepublic noted that interactive tools still need to be checked for figures, calculations, and assumptions, especially when used to compare costs or make business decisions. An easy-to-use interface does not inherently validate that the underlying data is accurate.