Runway has announced the deployment of its Gen-4.5 video generation model and the new GWM-1 general world model family on NVIDIA’s Vera Rubin NVL72 platform. These advancements mark Runway as one of the first to demonstrate high-performance, long-context video generation capabilities on NVIDIA's latest AI infrastructure. Gen-4.5, regarded as the leading video generation model globally, was transitioned from the previous NVIDIA Hopper architecture to Rubin within a single day, showcasing full compatibility and readiness for large-scale production workloads.
The Gen-4.5 model is designed for real-time, high-fidelity video generation, leveraging 50 petaflops of inference compute per GPU to deliver longer and more complex video sequences. Runway’s GWM-1 introduces real-time, physics-aware simulations for applications such as:
- Robotics training
- Explorable virtual environments
- Interactive avatars
This expands the scope of generative AI beyond static images or text-based models. Both models are now available to enterprise and research partners, with the Rubin platform supporting global deployment for users in entertainment, robotics, and virtual reality sectors.
Runway, a leader in creative AI tools, has partnered closely with NVIDIA to push the boundaries of generative AI, focusing on infrastructure that supports demanding computational workloads. The company’s collaboration with NVIDIA reflects a strategic commitment to making advanced video and world simulation technologies accessible to industry professionals and researchers, enabling broader experimentation and practical deployment of AI systems that can understand and simulate the physical world.
