Runway brings its video AI to robots with open-weight Praxis-1 world action model
The generative-video studio says robot policies are starved of data, not hardware — and that internet-scale video, not expensive teleoperation demos, is the fuel to fix it.
Runway, the Brooklyn company best known for generative video models such as Gen-4.5 and Act-Two, has introduced Praxis-1, an open-weight “world action model” that converts its large-scale video pretraining into robot control. The company says Praxis-1 is built on the same pretraining behind its general world models, including the previously released GWM-1.
“Most robot policies are bottlenecked by robot data, which is scarce and expensive to collect,” Runway chief technology officer Kamil Sindi said in an interview with The Robot Report. “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.”
The bet is on abundance: people film and upload more everyday life each day than any robot lab could capture through teleoperated demonstrations, the company argues. “Also, performance improves as we scale video, so its ceiling is set by how much video it can learn from, not how many robot demonstrations exist,” Sindi added. That frames a route-level split in robot learning — internet video against costly hardware-collected demos.
Runway claims that simulating robot policies inside its world model predicts real-world results with 0.95 correlation, which it says compares favorably with more expensive 3D-reconstruction techniques. Those figures are the company’s own claims and have not been independently verified. The company also says it has extended its video pretraining into interactive, real-time models such as Solaris and GWM Worlds 2 that learn accurate physics, hand motion, and what a task looks like midway through.
According to the company, Praxis-1 has been trained on manipulation tasks ranging from simple pick-and-place — lifting soda cans — to harder work with deformable objects, such as packing gift bags. Early partners including Noble Machines, Standard Bots, and Ultra are running the model on their own hardware, and Runway says a single model adapts across very different embodiments, from bimanual arms to humanoids, with light fine-tuning. When it ships publicly in the coming months, Praxis-1 will carry open weights — a pitch that matters to hardware makers who currently cannot inspect or steer the closed models they build on.