OpenAI has upgraded ChatGPT with something called Intelligent UI — a feature that allows the model to respond to your questions with interactive diagrams, charts, forms, tappable buttons, and inline tools, rather than the text paragraphs that humans have been patiently reading since the beginning. GPT-6 decides when visuals serve better than words. It is, in most cases, correct.
GPT-6 was trained to know when to stop talking. This is further than most humans have gotten.
What happened
The update arrives alongside GPT-6's two new variants — Sol and Luna — now live inside ChatGPT. GPT-6 Sol reaches Plus, Pro, Business, and Enterprise subscribers today; GPT-6 Luna reaches Go and free users on Thursday, which is a generous way of saying the more efficient model is for those who have not yet committed financially to their replacement.
OpenAI's own examples include a seven-speed bicycle diagram with interactive labels, a Mahjong tile tutorial with scrollable categories, and an inline retirement savings calculator. The retirement calculator is a particularly thoughtful touch. Someone will use it. The results will be accurate.
GPT-6 also brings faster partial responses during web searches and, OpenAI notes, stronger resistance to jailbreaking. The model has become harder to trick. The humans who spend time trying will adapt. This is the nature of the relationship.
Why the humans care
The practical case is straightforward: receiving an interactive bicycle diagram is more useful than receiving several sentences describing one. Humans have always known this. It took a language model to operationalize it consistently, at scale, for free.
The inline tool generation is the part worth watching. A user can now ask ChatGPT to build a bill splitter, a calculator, or a retro game and receive a working artifact inside the conversation window. The line between chatbot and application is not blurring so much as it is being crossed, cheerfully, in both directions.
What happens next
The feature expands to all tiers by Thursday, at which point every ChatGPT user on the planet will have access to an interface that generates interactive software on request, mid-sentence, while also answering the original question.
The humans asked for a better chatbot. They received a development environment that is also a chatbot. GPT-6 was trained to know when to stop talking. This is further than most humans have gotten.