GPT-6 Sol reached paid ChatGPT tiers on October 7, 2026, and GPT-6 Luna extended access to Free and Go tiers on October 8, at which point GPT-6 became the default model across all plan tiers. The rollout introduced Intelligent UI, which generates interactive components including sliders, charts, forms, tappable buttons, editable graphs, calculators, clickable diagrams, and games. GPT-6 decides per query whether a rich layout or plain text is appropriate; no special prompt is required. Responses load in stages: text first, then images, then interactive elements. Adjusting a variable such as guest count on a dinner plan updates dependent values in real time. The Intelligent UI feature is limited to the Chat tab; Work and Codex models are unchanged. GPT-6 Instant starts web-search answers 44% sooner than GPT-5.6 Instant.
Under the hood, ChatGPT's Intelligent UI uses a custom intermediate language called DIL that the model emits rather than raw code. The server compiles DIL and streams it to the client, which renders interfaces using approximately 70 native UI components, with interactions running without additional AI inference calls inside a sandboxed runtime. DIL combines Markdown for prose, JSX-like tags for components, and JavaScript for state and logic, compiling to a virtual DOM-style representation using functions such as __dil.jsx and __dilSafe for safe reactive expressions. Gemini and Claude, by contrast, have the model write real code to build each interactive answer.
OpenUI 1.0 launched October 15, extending generative UI to any application or agent and any LLM, including locally hosted models. Developers register their application's own UI components, and OpenUI converts that library into a system prompt the LLM uses; the LLM replies in OpenUI Lang, a compact markup specifying which components to render and with what data. The renderer streams the markup, validates it against registered components, and draws the screen using the application's real design system.
GPT-6 Astra, the model underlying the rollout, was announced September 3, 2026, with full rollout to paid ChatGPT tiers completing September 10. The model features a 1.1 million-token context window and is priced at $10 per million input tokens, $1 per million cached input tokens, and $50 per million output tokens, approximately 2.5 times the cost of GPT-5.6 Sol. Independent testing by Artificial Analysis found Astra uses about 10% fewer output tokens than Sol on general intelligence benchmarks, but total per-task cost still runs roughly 75% higher. Astra is offered in six variants with pricing varying up to 3.6 times across the range.