Runway unveils GWM-1, a General World Model family designed to unify multiple domains under one base model and extend capabilities beyond a single general model.

GWM-1 is a real-time, simulating world model that users can interact with through actions like camera pose, robot commands, and audio inputs.

Under the GWM umbrella, Runway highlights three facets: GWM-Worlds for interactive scene generation at 24 fps 720p, GWM-Robotics for synthetic data with variable weather and obstacles, and GWM-Avatars for realistic human-like conversational agents.

The piece includes credits to TechCrunch reporters and provides contact details for more information.

Use cases highlighted include gaming, education, training agents, and immersive experiences, emphasizing coherent, real-time simulation for training and evaluation.

A strategic partnership with CoreWeave will run Runway on CoreWeave’s cloud, using Nvidia GB300 NVL72 racks to scale training and inference.

Runway differentiates with industry credibility and user-focused design, though faces a crowded world-model space where some competitors have more resources.

The livestream underscores commercialization and professional adoption prospects across film/TV, advertising, game development, robotics, and scientific research, signaling broad real-world impact ambitions.

Gen 4.5 updates will roll out first to enterprise customers, then to all paid plan users in the coming weeks.

The coverage places Runway within the broader world-model race among AI image/video companies and tech giants, noting related industry developments and rivals.

GWM-Robotics and GWM-Avatars are offered via an SDK, with ongoing discussions with robotics firms and enterprises to deploy these capabilities.

GWM-1 is positioned as more general than competitors like Google’s Genie-3, with potential uses in training agents for robotics and life sciences.