ModelsNews
Runway Praxis-1: open-weight robotics model under Apache 2.0
Runway released its Praxis-1 world action model with open weights under Apache 2.0. The 27-billion-parameter MoE diffusion model leads NVIDIA's RoboLab-120 leaderboard at 46.2% success and can be adapted to your own robot with LoRA.
· 2 min read

What Runway released
Praxis-1 was introduced in September as Runway's first world action model. Now anyone can download the weights plus the inference and fine-tuning code under Apache 2.0, including for commercial use (Runway).
- A 27B-parameter mixture-of-experts diffusion transformer that predicts video and actions together.
- Built on the same large-scale video pretraining as Runway's general world models.
- Available on GitHub, on Hugging Face with native LeRobot support, in vLLM-Omni, and with hosted inference on Reactor.
- Runs on H100 and B200, RTX 5090 and RTX PRO 6000, and on Jetson Thor on the robot itself.
The numbers
On NVIDIA's RoboLab-120 leaderboard it reaches 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 outperforms every open-weight VLA.
Fine-tuned on Google's RT-1 robot, which never appeared in its training data, Praxis-1 reaches 80.0% on SimplerEnv, against 52.4% for the original RT-1 policy.
Adapting it to your hardware
Launch partners Noble Machines, Standard Bots and Ultra ran it on a legged humanoid, a six-axis industrial arm and a dual-arm warehouse robot. Both full fine-tuning and LoRA are supported, and LoRA lets you adapt it to your embodiment, cameras and tasks without retraining the full model. NeMo Automodel tutorials cover post-training.
Why it matters
A commercially usable open-weight robotics model at this level is rare. Runway is a founding member of the Cosmos Coalition, and we covered the Cosmos 3 models earlier (NVIDIA Cosmos 3: four open world simulation models). Next to the closed Gemini Robotics 2.5, Praxis-1 is the option a university lab or startup can actually download and test on its own robot today.
Produced with the help of AI, reviewed by an editor. How we make our articles
Reviewed and edited by the Neural Daily desk. About


