FreshSynapse

AI trends & daily updates, in plain English

AI Models

Runway Gives Away Its Robot Brain: Praxis-1's Weights Are Now Open, Apache 2.0, Commercial Use Included

By FreshSynapse Editorial · · 4 min read

A large orange industrial robotic arm in a workshop, with a smaller white robot arm on a workbench behind it

Photo: Hippopx (public domain). Illustrative image.

What happened. Runway released the weights for Praxis-1 on Tuesday 11 November 2026, giving away the robot model it first showed in September. The release is unusually complete for an open model: the weights, inference code and both full and LoRA fine-tuning code are published under the Apache 2.0 licence, which permits commercial use, on GitHub and Hugging Face, with native support for the LeRobot robotics library and serving through vLLM-Omni. Hosted inference is available on Runway’s Reactor platform. Anyone can download it, run it on their own hardware, fine-tune it for their own robot, and use it commercially — no licence negotiation, no API gatekeeper.

What Praxis-1 is. Runway calls it a world action model: a 27-billion-parameter mixture-of-experts diffusion transformer that predicts video and robot actions together. The mechanism is the interesting part. Instead of mapping camera images directly to motor commands, the system predicts a short video clip of what should happen next, and derives the robot’s actions from that imagined future. Each call to the model produces 1.6 seconds of motion — 24 action rows at 15 Hz — in roughly half a second on a single H100 GPU, according to Runway’s technical write-up. It learned to move by watching video first, then was fine-tuned on robot data. That video-first training is the same idea behind world models that learn how reality behaves — here applied to hands, not frames.

The benchmark numbers. Runway’s headline number comes from RoboLab-120, Nvidia’s simulated benchmark of 120 tabletop manipulation tasks with 10 trials each. The DROID fine-tuned version of Praxis-1 completed 555 of 1,200 episodes, a 46.2% success rate, ahead of FLUX 3 Action at 42.9%. Runway says that is the best result among the listed world action models and open-source robot foundation models. Fine-tuned on Google’s RT-1 robot, which it never saw in training, it reaches 80.0% on SimplerEnv, against 52.4% for the original RT-1 policy. Two caveats belong next to those numbers. First, every figure so far is Runway’s own — no independent lab has reproduced the leaderboard result yet. Second, 46.2% is a first-place finish that still means failing more than half the time, on simulated tabletop tasks, not a factory floor.

What you can download. Three checkpoints are available: the base mid-trained model, a version fine-tuned on the DROID robot dataset, and one fine-tuned for the bimanual YAM robot. The full model is 54.4 GB in bfloat16 and needs a single 80 GB GPU; a 4-bit NVFP4 quantised version shrinks the weights to 17.5 GB. Runway says it runs out of the box across Nvidia hardware, from H100 and B200 data-centre cards to RTX 5090 and RTX PRO 6000 workstation cards, and on the Jetson Thor computer that sits inside the robot itself. Three launch partners have already run it on real machines: Noble Machines on a legged humanoid, Standard Bots on a six-axis industrial arm, and Ultra on a dual-arm warehouse robot.

Why it matters. Praxis-1 arrives in the middle of a crowded week for robot models. Nvidia opened up its GR00T 2 model days earlier, and Google’s Gemini Robotics 2.5 is already running on a Mercedes-Benz factory line. Runway’s entry is the most permissive of the three: full weights, fine-tuning code, and a licence that allows commercial use. That matters because robotics has been bottlenecked on data — every lab collecting its own robot demonstrations, slowly. A video-trained model that anyone can adapt could break that bottleneck, the way open language models did for text. The warning stands alongside the promise: open model weights do not automatically produce safe robot motion. A model that predicts plausible actions can still crush, drop or misjudge, and the safety work now falls to whoever deploys it.

Sources

  • Runway
  • Praxis-1
  • robotics
  • open weights
  • AI models
  • world models

Related articles