Industry
Runway releases Praxis-1, an open-weight world action model for robots
Runway said on 11 November that Praxis-1, the world action model it introduced in September, is now available to developers and researchers with open weights under the Apache 2.0 licence. In a company post, Runway said launch partners including Noble Machines, Standard Bots and Ultra have already been running the model on their own hardware: a legged humanoid, a six-axis industrial arm and a dual-arm warehouse robot.
The company describes Praxis-1 as a 27-billion-parameter mixture-of-experts diffusion transformer that predicts video and actions together, built on the same large-scale video pretraining as its general world models. Runway said it is the top-performing world action model on NVIDIA's RoboLab-120 leaderboard and that it outperforms every open-weight vision-language-action model in that comparison. On RoboLab-120, the post lists a 46.2% success rate across 1,200 episodes, compared with 42.9% for FLUX 3 Action, 36.8% for Cosmos3 Nano and 28.0% for π0.5, with the figures given as of November 2026. Those rankings are Runway's account of the benchmark.
The weights, inference code and fine-tuning code, including full fine-tuning and LoRA, are released under Apache 2.0 for commercial use, the post says. Praxis-1 is available on GitHub, on Hugging Face with native LeRobot support, and in vLLM-Omni, with hosted inference on Runway's Reactor service. Runway said it runs on NVIDIA hardware from data-centre GPUs such as the H100 and B200, through workstation cards such as the RTX 5090 and RTX PRO 6000, to Jetson Thor on a robot.
Runway said a single Praxis-1 checkpoint has been adapted to different robot bodies with light fine-tuning. Fine-tuned on Google's RT-1 robot, which it said never appeared in the model's training data, Praxis-1 reached 80.0% success on SimplerEnv, above the original RT-1 policy at 52.4%. The company also said its research found that web video matches teleoperated robot video as pretraining data for policy performance. Those results are Runway's.
The post ties the release to Runway's role as a founding member of the Cosmos Coalition with NVIDIA. Developers can download the model on Hugging Face, view the code on GitHub, try it on Runway Dev, or post-train it with NeMo Automodel tutorials, the company said.