Introducing Solaris
By Steven Van ·
Runway introduced Solaris, the first of its Interface World Models.
Runway has introduced Solaris, the first of what it calls Interface World Models: a real-time model that generates an interface's frames, and its response to every click or drag, live as someone uses it, instead of running on pre-built code.
Runway describes three capabilities this enables:
- The interface is entirely visual, with no separate coded layer behind what's shown, so a scene like a virtual clothing rack can be manipulated directly rather than clicked through menus.
- It stays "alive", rendering continuously rather than waiting for the next action, so details like reflections or object positions can shift on their own.
- It is open-ended, able to support behaviors that weren't explicitly built in, instead of being limited to interactions a developer anticipated.
Solaris builds on Runway's Gen-4.5 video model and follows its earlier GWM-1 general world model. In its own evaluations, Runway says it was preferred over coded interfaces 61% of the time on instruction following and 71% on natural behavior. It is not a public product: Runway is launching with select partners and taking early-access requests, and lists legible text, coherence over long sessions, and screen-reader support as open problems.
