Runway has introduced its first world model, GWM-1, marking a shift from its focus on video generation. This model consists of three autoregressive systems built on the Gen-4.5 text-to-video generation framework. Each model is tailored with specific domain data to serve various applications. The GWM Worlds model allows users to interact with and shape digital environments in real time. It can maintain coherence over lengthy sequences, making it useful for tasks like game design, virtual reality, and educational explorations. While it may not be a full simulation, it’s reliable enough for practical use. The second component, GWM Robotics, focuses on generating synthetic training data for robotics. It enhances existing datasets by introducing new objects, task instructions, and variations in environments. This ability to expand training data is crucial for improving AI agents and robots, providing diverse scenarios that can enhance their training. Both models represent a strategic move for Runway as it ventures beyond its traditional video generation offerings into broader applications in AI and robotics.