OpenAI has begun rolling out a new Intelligent UI capability in ChatGPT, available to every user as part of the GPT-6 release. The feature lets the assistant embed visual elements such as diagrams, charts, forms and clickable buttons directly within its textual replies, turning conversations into richer, more interactive experiences.
In a blog post, OpenAI explained that GPT-6 was trained to determine when an answer benefits from a visual supplement and how to format that element correctly. One illustration provided shows a schematic of a seven-speed bicycle, where users can tap buttons to highlight individual components such as the chain, gears or frame.
Another demo demonstrates how the system can teach a game like Mahjong by presenting each tile as an image that can be scrolled through and sorted into categories. The same interface can also generate on-the-fly tools, for example a retirement-savings calculator, a retro-style video game, or a bill-splitting utility, all embedded in the chat window.
OpenAI introduced the first GPT-6 models earlier this year and now adds the Sol and Luna variants to the ChatGPT platform. Alongside the visual upgrades, the new models are reported to perform faster web searches and to provide provisional answers while the system continues gathering information, reducing latency for user queries.
The company also notes that GPT-6 displayed stronger resistance to attempts at bypassing its safety training, suggesting an improvement in alignment and robustness. No further quantitative metrics were disclosed, but the claim points to ongoing efforts to mitigate misuse as the model’s capabilities expand.