Runway Launches Solaris, First 'Interface World Model,' Beats Claude in Blind Tests
新智元 · wechat · 2026-09-01
Runway has released Solaris, the first "interface world model": no code required, it generates interactive UIs frame-by-frame in real time, with clicks and drags conditioning the next frame. Built on the Gen-4.5 video model and continuing the GWM-1 world-model line, it combines autoregressive generation, diffusion distillation, and training on its own outputs for stability; an LLM handles reasoning while Solaris renders.
In blind tests with 250 participants and 7,500 pairwise comparisons against Claude Opus 5's code-generated interfaces, Solaris won instruction following 61% vs 24% and natural fluidity 71% vs 21%. Runway admits weaknesses in text clarity, hallucination, and long-horizon coherence. Co-founder Cristóbal Valenzuela declares "video is the universal interface" — the first step toward all UIs becoming chat+gesture input with video output, and a new way to generate dynamic environments for agent training.
Related event: Runway Unveils Solaris, Its First Interface World Model(16 posts)→
More from Multimodal
- FLUX.2 Klein Drifts Hard on Character Expressions While Free Gemini Holds Likeness — wacomlover · 2026-09-11
- Tencent Hunyuan releases AuK code and weights on GitHub with ComfyUI and fine-tuning support — aigclink · 2026-09-11
- Creator turns Bahamut vs Tiamat rivalry into an AI cinematic battle with Midjourney, GPT Image 2 and Seedance — azed_ai · 2026-09-11
- invideo launches AI agent-powered editor to automate repetitive editing tasks — azed_ai · 2026-09-11
- fable 5.1 recreates The Starry Night with 256,157 JavaScript brush strokes — cedric_chee · 2026-09-11
- GPT-6 Astra + Hyper3D Rodin MCP Generates 3D Assets in One Agent Flow — ahuja_priyank · 2026-09-11