Runway has upgraded Gen-4.5 and introduced GWM-1, the company's first "General World Model."
The recently introduced Gen-4.5 now features native audio generation and audio editing, as well as multi-shot editing, a feature that lets users apply changes to a single scene and have them ripple across the entire video.
The new GWM-1 builds an internal representation of an environment to simulate future events in real time. Built on the Gen-4.5 architecture, it generates video frame by frame while allowing for interactive control through inputs like camera movements, robot commands, or audio.
The world model ships in three distinct versions: GWM Worlds for creating explorable environments, GWM Avatars for generating speaking characters with realistic facial expressions and lip sync, and GWM Robotics for producing synthetic training data for robots. Runway plans to eventually merge these capabilities into a single unified model.
AI labs race to build world models
Runway isn't the only one chasing this technology. The field is becoming crowded.
Other labs, including Google DeepMind and a new startup from AI researcher Yann LeCun, are also developing world models. The industry views these systems as a critical evolution beyond conventional language models, which still lack a fundamental understanding of the physical world. Google DeepMind CEO Demis Hassabis confirmed that building these models is central to the company's strategy for reaching Artificial General Intelligence (AGI).
The race also includes World Labs, a startup founded by Fei-Fei Li that raised $230 million to develop "Large World Models" (LWMs) with spatial intelligence. The company recently unveiled "Marble," a prototype capable of rendering persistent 3D environments from multimodal prompts. Munich-based startup Spaitial is developing Spatial Foundation Models designed to generate and interpret 3D worlds with consistent physical dynamics.
The competition continues to expand. Startups Etched and Decart recently introduced the "Oasis" project, a system that generates playable, Minecraft-style 3D worlds in real time at 20 frames per second. While it allows basic interactions like jumping and picking up objects, it still faces consistency challenges—players sometimes find themselves in different environments simply by turning around.
In August, Chinese tech giant Tencent released its Hunyuan World Model 1.0, an open-source generative AI model that creates 3D virtual scenes from text or image prompts.
