Runway opens Praxis-1 weights

Runway's post on 11 November 2026 says Praxis-1, its world action model, is available with open weights under the Apache 2.0 license, including commercial use. Runway introduced it in September. Launch partners named on the post, Noble Machines, Standard Bots, and Ultra, have been running it on a legged humanoid, a six-axis industrial arm, and a dual-arm warehouse robot. The weights, inference code, and fine-tuning code, including LoRA, are on GitHub and Hugging Face, with native LeRobot support, vLLM-Omni, and hosted inference on Reactor.

Runway describes Praxis-1 as a 27-billion-parameter mixture-of-experts diffusion transformer that predicts video and actions together. On NVIDIA's RoboLab-120 leaderboard, Runway says it reaches 46.2% success across 1,200 episodes, against 42.9% for FLUX 3 Action, 36.8% for Cosmos3 Nano, and 28.0% for π0.5. Runway dates that table to November 2026. Fine-tuned on Google's RT-1 robot, which Runway says never appeared in the training data, Praxis-1 reaches 80.0% success on SimplerEnv, against 52.4% for the original RT-1 policy. Runway says it runs on NVIDIA hardware from H100 and B200 GPUs to an RTX 5090, an RTX PRO 6000, and Jetson Thor on the robot.

Who it's for: a robotics developer who wants to run or fine-tune an open world action model on their own hardware.

Fits: Best for Video stays Runway, for camera control, inpainting, and a workflow that resembles editing. Praxis-1 predicts a robot action. It leaves that editing job where it is. The video guide and the Runway alternatives are unchanged. The verdict is unchanged.