AI image and video company Runway launched its first world model, GWM-1, designed to simulate real-world physics and behavior frame by frame. World models allow AI to reason, plan, and act without needing training on every scenario, enabling applications in robotics, life sciences, and interactive projects. Runway says GWM-1 is more general than competitors, including Google’s Genie-3, and emphasizes its ability to create realistic simulations that reflect geometry, lighting, and physical dynamics.
Runway introduced specialized applications for the model. GWM-Worlds allows users to generate interactive environments for gaming or agent training, dynamically producing scenes as users explore them. GWM-Robotics focuses on synthetic data generation, simulating variable conditions such as weather or obstacles to test robotic systems and detect policy violations. GWM-Avatars is designed to create realistic human simulations for training, communication, and other applications, competing with companies like D-ID, Synthesia, and Soul Machines.
The launch accompanies an update to Runway’s Gen 4.5 video model, which adds native audio, multi-shot video generation, and one-minute video production with consistent characters, dialogue, and complex scenes. The Gen 4.5 update will roll out first to enterprise customers and later to all paid users. With GWM-1 and the Gen 4.5 improvements, Runway positions itself at the forefront of production-ready AI video and simulation tools for commercial and research applications.
