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

Runway, the New York company best known for AI video generation, has released Praxis-1, its world action model for robotics, to developers with open weights. According to The Robot Report, the model is available under the Apache 2.0 license, including for commercial use, along with inference and fine-tuning code.

What the model does

Praxis-1 is a 27-billion-parameter mixture-of-experts diffusion transformer that predicts video and actions together. Built on Runway's video foundation model, it pictures what the scene should look like next and produces the motion to get there. "A model trained to predict video really understands a lot about the physical world," said Andy Chen, Runway's head of robotics.

On NVIDIA's RoboLab-120 benchmark, Praxis-1 reached 46.2% success across 1,200 episodes, ahead of FLUX 3 Action (42.9%), Cosmos3 Nano (36.8%) and π0.5 (28.0%). Runway says it is the top-ranked open-source, end-to-end policy on the leaderboard.

Partner deployments

Runway announced Praxis-1 in September with launch partners Noble Machines, Standard Bots and Ultra, which have run it on a legged humanoid, a six-axis industrial arm and a dual-arm warehouse station, with the same checkpoint adapted through light fine-tuning. Chen said that with Noble Machines, Praxis-1 learned a warehouse plank-picking task from half an hour of examples and succeeded 15 times out of 15.

The model runs on NVIDIA hardware from H100 and B200 data center GPUs to RTX 5090 workstations and Jetson Thor on the robot itself, and is available on GitHub, Hugging Face and vLLM-Omni.

Why open weights

Runway says open weights give hardware developers the flexibility to run, inspect and adapt a world action model on their own terms. The company now plans to focus on scaling real-world deployments in warehouses, factories and industrial sites with its partners.

Sources: The Robot Report