Loading page…
Intelligent UI is designed to make complex topics easier to learn. Users can reduce the visuals, and free and Go accounts follow the initial rollout by one day.
Listen to this story
The update shifts ChatGPT answers from mostly text toward task-specific interfaces that users can manipulate, including editable graphs and calculators embedded in responses. OpenAI says the goal is to support learning and everyday planning, while letting users reduce visual elements if they prefer. The staged rollout reaches paid plans first on October 7 and Free and Go users on October 8, giving OpenAI a way to test a more interactive format across its user tiers.
Demonstrations included an airplane-wing explanation, recipe visuals, bicycle mechanics, a hiking map and a personal savings calculator.
Pro, Plus, Business and Enterprise users are scheduled to get Intelligent UI on October 7; Free and Go users follow October 8.
TechCrunch reports the rollout will arrive alongside GPT-6; the article does not specify a year for the October dates.
ChatGPT answers are gaining elements people can work with directly: editable graphs, interactive charts and calculators built for a particular task. OpenAI is rolling out the feature, called Intelligent UI, starting October 7. The company's aim is to make complex subjects easier to learn—not simply to add more pictures to a conversation.
At a demonstration for journalists, OpenAI product manager Aarush Selvan described the starting point as a ChatGPT interface that has predominantly relied on text. He framed the change around everyday requests, from planning a recipe or a trip to learning something new, according to TechCrunch.
The new interface adds more than static illustrations. OpenAI says responses can contain tappable buttons, task-specific calculators, interactive charts and editable graphs. Those elements put controls inside the answer itself: a graph can be edited, rather than only viewed, and a calculator can be tailored to the task under discussion.
Selvan used a question about flight to show the intended learning experience. He asked ChatGPT to explore how an airplane wing generates lift. The chatbot produced several diagrams designed to explain how planes stay in the air, giving the explanation a visual form alongside the conversation.
OpenAI's other examples extended beyond that classroom-style explanation. The company showed or described recipe visuals, a diagram of bicycle mechanics, a map for a multi-day hiking excursion and a personal savings calculator. The examples span both explaining a subject and helping someone plan an everyday activity.
The global rollout starts Wednesday, October 7, for Pro, Plus, Business and Enterprise users, alongside GPT-6, TechCrunch reports. Free users and subscribers to the lower-cost Go tier are scheduled to receive Intelligent UI on Thursday, October 8. The schedule puts the interface on both paid and free accounts, with a one-day gap between the groups.
Story updates
OpenAI began rolling out GPT-6 globally on October 7, 2026, bringing interactive answers to everyday ChatGPT conversations across free and paid plans. The models can also begin answering before they finish reasoning. But OpenAI’s accompanying safety assessment shows mixed results: stronger defenses against manipulation alongside setbacks in several tests, including those designed to assess protections for teenagers.
The rollout began October 7 for Plus, Pro, Business and Enterprise, with Free and Go scheduled to follow October 8. Enterprise access depends on administrator settings. GPT-6 Sol powers the four higher tiers; GPT-6 Luna serves Free and Go.
Sol and Luna replace their GPT-5.6 counterparts in Chat. Work and Codex retain previously released September versions of GPT-6, distinct from the October models assessed for this launch.
In its launch announcement, OpenAI describes Intelligent UI: answers combining text with graphics, buttons, forms and interactive tools. A library of interface components supplies the building blocks. A compiler processes the interface during generation, letting elements appear progressively rather than waiting for a finished response.
OpenAI’s system card reports stronger resistance to jailbreaks—attempts to bypass safety rules—including attacks that adapt over several conversation turns. It also reports less dishonesty, deception and guardrail circumvention than the GPT-5.6 models being replaced.
Sol showed a statistically significant regression on the standard self-harm evaluation. Luna regressed on standard self-harm, gore and sexual content. OpenAI says manual review found borderline violations that were generally safe, while adversarial testing identified no high-severity risks under its criteria.
The company also reports fewer refusals of harmless requests and fewer excessive or judgmental caveats, despite lower scores on some safety requests.
Both October models regressed significantly on age-restricted content, sexual content and emotional reliance compared with their respective GPT-5.6 August versions.
Luna also regressed significantly on gore. Other differences in the under-18 evaluations were not statistically significant.
OpenAI applies an additional classifier-based block to teen responses that may contain self-harm, sexual content or gore. That protection is not included in the reported evaluation scores. For emotional reliance, the company says its test is overly sensitive to benign nicknames such as “bro” or “bestie,” which are allowed when explicitly requested.
openai.comLoading discussion...
Join the conversation
Share a task where you would welcome visuals or prefer plain text.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.