Details
- Runway announced Praxis-1, its first open-weight world action model for controlling real robots.
- The model applies Runway’s large-scale video pretraining to robotics, converting visual knowledge about physical tasks into robot control.
- Runway positions Praxis-1 as a generalist policy model designed to operate across different robot embodiments and environments.
- The company is testing Praxis-1 with early partners Noble Machines, Standard Bots, and Ultra on their own hardware.
- Praxis-1 is not yet publicly downloadable; Runway plans to release its weights publicly in the coming months.
- Runway says the approach addresses a key robotics constraint: robot demonstrations are scarce and expensive, while video data is widely available.
- Open weights could allow robotics developers and researchers to inspect, adapt, fine-tune, and run the model on their own systems once released.
Impact
Praxis-1 extends competition in embodied AI beyond companies built around robotics data by applying a video-modeling pipeline to physical control. Its open-weight plan could lower experimentation costs and broaden access for robotics researchers, although the model’s practical significance will depend on transfer reliability across hardware, safety performance, licensing, and the results of partner testing. The delayed public release also means its near-term impact remains limited to selected deployments.
