Runway Releases Praxis-1 Open-Weight World Action Model for Robotics

Runway AI Inc. released Praxis-1, a world action model for robotics, with open weights under Apache 2.0. Developers can now download and adapt the model, first announced in September with Noble Machines, Standard Bots and Ultra.

Those partners have tested it on a legged humanoid, a six-axis industrial arm and a dual-arm warehouse robot.

Andy Chen, Runway's head of robotics, told The Robot Report that a strong base model makes a major difference. He said Runway's earlier model needed extensive training with Noble Machines for a warehouse task: picking up wooden planks one at a time and placing them in a container.

Praxis-1, trained with more compute and data, learned it from half an hour of examples and succeeded 15 times out of 15. Given a messy pile of unseen planks to clear into bins, it cleared the whole pile six times out of 10.

Chen said partner feedback improves Runway's models.

He noted that Praxis-1 is the top-ranked world action model on NVIDIA's RoboLab leaderboard and the top-ranked open-source, end-to-end policy.

Praxis-1 is on GitHub, Hugging Face with native LeRobot support, and vLLM-Omni, with hosted inference on Reactor. Runway released weights plus inference and fine-tuning code under Apache 2.0, including for commercial use.

The release supports full fine-tuning and low-rank adaptation (LoRA). The model runs out of the box on NVIDIA hardware, from H100 and B200 data-center GPUs to RTX 5090 and RTX PRO 6000 workstation GPUs to Jetson Thor on the robot.