Runway launches GWM-1 world model, expanding beyond video generation into robotics and AI agents

Runway Unveils GWM-1 World Model for AI Agent Training

Runway has announced GWM-1, its first world model that marks a strategic pivot beyond the company's established reputation in AI-driven video generation 1

1

. The system works through frame-by-frame prediction to create a simulation with an understanding of physics and how the world behaves over time 2

2

. This development positions Runway alongside an increasing number of startups and tech companies racing to build world models as large language models and video generation enter a refinement phase.

GWM-1 functions as a blanket term for three autoregression models, each built on top of Runway's Gen 4.5 text-to-video generation model and then post-trained with domain-specific data for different applications 1

1

. Runway is pitching it as more "general" than Google's Genie-3 and other competitors, emphasizing its capacity to create simulations for training AI agents and robots across domains like robotics and life sciences 2

2

.

GWM Worlds Enables Real-Time Digital Environment Exploration

GWM Worlds offers an interface for real-time digital environment exploration where user input affects the generation of coming frames 1

1

. Users can define the nature of the world, including what it contains and how it appears, as well as establish rules like physics. The system allows for actions or changes reflected in real-time, such as camera movements or descriptions of environmental modifications. Runway claims these simulations can remain consistent and coherent across long sequences of movement.

Potential applications extend to game design and development for pre-visualization and early iteration, generation of virtual reality environments, and educational explorations of historical spaces 1

1

. While the methodology relies on advanced frame prediction rather than full-on world simulations, Runway suggests the technology is reliable enough for practical use. The company noted that while GWM Worlds could prove useful for gaming, it's also well positioned to teach agents how to navigate and behave in the physical world 2

2

.

GWM Robotics Generates Synthetic Training Data for Robot Development

The second model variant, GWM Robotics, addresses a major use case that extends beyond Runway's traditional focus area 1

1

. The system generates synthetic training data that augments existing robotics datasets across multiple dimensions, including novel objects, task instructions, and environmental variations. Runway aims to use this synthetic data enriched with new parameters like changing weather conditions or obstacles 2

2

.

The company says this method could also reveal when and how robots might violate policies and instructions in different scenarios, offering a testing ground for safety and compliance before real-world deployment 2

2

. Runway confirmed it will make GWM Robotics available through an SDK and is in active conversations with several robotics firms and enterprises for deployment.

GWM-Avatars Simulates Human Behavior for Communication Applications

Runway is also building realistic avatars under GWM-Avatars to simulate human behavior 2

2

. This positions the company in competition with established players like D-ID, Synthesia, Soul Machines, and Google, who have worked on creating human avatars that appear realistic and function in areas like communication and training. The company is currently in active discussions with enterprises regarding the deployment of GWM-Avatars alongside GWM Robotics.

Gen 4.5 Update Brings Native Audio and Long-Form Multi-Shot Generation

Beyond releasing its world model, Runway is updating its foundational Gen 4.5 model that launched earlier this month and surpassed both Google and OpenAI on the Video Arena leaderboard 2

2

. The new update introduces native audio and long-form multi-shot generation capabilities. Users can now generate one-minute videos with character consistency, native dialogue, background audio, and complex shots from various angles.

This update nudges Runway closer to competitor Kling's all-in-one video suite, which also launched earlier this month, particularly around native audio and multi-shot storytelling 2

2

. The enhanced Gen 4.5 model will be available to enterprise customers first before rolling out to all paid plan users in the coming weeks, signaling that video generation models are moving from prototype to production-ready tools.