On August 31, 2026, Runway unveiled Solaris, the first model in what the company calls a new category: "Interface World Models." Instead of generating code that later gets rendered, Solaris draws the interface directly, frame by frame, responding in real time to the user's clicks and drags — with no intermediate representation like HTML or a DOM in between.
According to Runway's official announcement and coverage from The New Stack, Solaris is a real-time world model that generates an interface frame by frame, learning from the user's clicks, drags, and other interactions to decide what to render next. A single world model produces every frame and every response to user input; a paired language model decides what the interface should do next, and the world model draws the result. There's no code layer underneath — no HTML, no DOM tree — the interface itself is live-generated video.
Solaris runs on Runway's Gen-4.5 video model, targeting 720p output with per-frame latency under 500 milliseconds, according to specs the company published. In head-to-head comparisons against interfaces built with traditional code, Runway reports Solaris was preferred 61% of the time on instruction-following and 71% of the time on natural behavior — figures that come from the company's own evaluation, with no independent audit available yet.
It's important not to oversell this: Solaris isn't launching as a public product. Runway says it's "working with key partners to bring Solaris to launch publicly" and is collecting requests through an early-access form, with no launch date or pricing announced. It's a category demo, not something a company can hire today.
Carlos Montiel is an enterprise AI solutions architect. He implements LLMs, Agents, RAG, and orchestrators for companies in Guatemala and Latin America. Reach out for a consultation.
Contact Carlos Montiel