Runway launches Praxis-1 open-weight world action model for robotics

Runway AI Inc. this week introduced Praxis-1, an open-weight world action model that converts its video pretraining into robot control. The company built Praxis-1 on the same large-scale video pretraining that supports its general world models. Runway said it is testing Praxis-1 across a variety of embodiments and environments to identify and close potential gaps before general availability.

The model is designed to give robotics developers and researchers a generalist policy that works across embodiments and environments. Runway plans to release Praxis-1 publicly in the coming months after early partner testing.

Kamil Sindi, Runway's chief technology officer, told The Robot Report that most robot policies are bottlenecked by robot data, which is scarce and expensive to collect. He said Praxis-1 learns mostly from third-person video, built on the same large-scale pretraining behind Runway's world models, so it already understands how objects behave and how tasks unfold.

Sindi added that performance improves as Runway scales video. Its ceiling is therefore set by how much video it can learn from, not by how many robot demonstrations exist, he said.

Founded in 2018, Runway AI specializes in generative artificial intelligence research and technologies. It offers models including Aleph 2.0 for in-context video editing, Act-Two for motion capture, and Gen-4.5 for text-to-video and image-to-video generation.

For robotics, Runway also offers GWM-1, a general world model. The Brooklyn, N.Y.-based company has offices in New York, San Francisco, Seattle, London, Paris, Tel Aviv, and Tokyo.

Runway is betting on video data for robot training. While real-world robotics data for generalist AI models is scarce, video data is abundant. Runway AI noted that people film and upload more of everyday life each day than any robot lab could capture through teleoperated demonstrations.

Runway said simulating robot policies inside its world model predicts real-world results with 0.95 correlation. The company claimed this compares favorably with more expensive 3D reconstruction-based techniques.

Runway has extended its work on pretraining large video models into interactive, real-time video models such as Solaris and GWM Worlds 2. By teaching models to generate accurate physics, how hands move, and what a task looks like partway through, Runway said it has created dynamic, complex environments for agent training in digital and physical worlds.

Runway said Praxis-1 brings the same approach to robotics, providing a generalist policy model for developers and researchers. Sindi said Praxis-1 has been trained on a variety of manipulation tasks, from straightforward pick-and-place actions like lifting soda cans to more complex tasks involving deformable objects like packing gift bags.