Radar / AI robotics / Runway Praxis-1
Runway Praxis-1: open-weight robot model that learned to act from video
Runway released Praxis-1, a 27B-parameter world-action model, with open weights, inference and fine-tuning code under Apache 2.0.
Why it matters
The most permissive open robot model yet: full weights plus fine-tuning code, topping NVIDIA's RoboLab-120 open-model leaderboard at 46.2%.
What you could build with it
A robotics startup could fine-tune Praxis-1 on its own arm's demonstration data using the released LoRA code, turning a general video-trained model into a working pick-and-place skill for a warehouse cell without collecting thousands of teleoperated trajectories from scratch.
Does it hold up?
Usable for research and fine-tuning today: it runs out of the box on NVIDIA hardware from H100 down to Jetson Thor, with a 4-bit build at 17.5 GB. Real-world reliability is unproven — the headline benchmark is vendor-reported simulation, and 46% success still fails most tasks.
Built with Runway Praxis-1
- Runway's Praxis-1 Beats Every Open Robot Model at 46% Success RateAlphaSignal · Breakdown of the open-weight release, the three checkpoints (base, DROID-tuned, bimanual YAM), and the RoboLab-120 result versus FLUX 3 Action and pi0.5.
- Runway Praxis-1: free open-weight robot AI model releasedmadrobot · Reports the launch partners Noble Machines, Standard Bots and Ultra running Praxis-1 on their own hardware, plus the 4-bit 17.5 GB quantized build.
- Runway open-sources Praxis-1: the robot model that learned to act by watching videoAI Frontier Post · Analysis of what ships in the download and the caveats: all benchmark figures are Runway's own and 46.2% still means failing over half of simulated tabletop tasks.
- runwayml/praxis-1-ossgithub · 73 ★ · Official release repo with inference code and full plus LoRA fine-tuning code for Praxis-1.
- huggingface/lerobotgithub · 28,831 ★ · Robotics library with native LeRobot support for Praxis-1, so the model plugs into existing LeRobot training and deployment pipelines.
Learn more
First spotted on article: source.
More AI robotics
RuView: see people through walls with WiFi, no camerasOpen-source Rust system that turns ordinary WiFi signals into real-time pose estimation, vital-sign monitoring, and…robotics · JEV 0.76NVIDIA GR00T 2: a world-action robot model that learns from human videoNVIDIA's open frontier world-action model predicts future video and robot actions together, hitting 71% one-shot…robotics · JEV 0.66RoboParty RP1: a full-stack open-source humanoid robot for embodied AI researchRoboParty's 1.3-metre humanoid, launched 16 October and back in coverage on 2 November, comes in 24-DoF standard and…robotics · JEV 0.62Gemini Robotics 2.5: DeepMind's most capable robot model, tested on a live Mercedes lineGoogle DeepMind introduced Gemini Robotics 2.5, its most capable vision-language-action model, reporting 30-100%…robotics · JEV 0.61T-Rex: a robot manipulation AI that reacts to fingertip touchPanasonic, UC Berkeley, NVIDIA and Stanford built T-Rex (Tactile-Reactive Dexterous Manipulation), a VLA model that…robotics · JEV 0.61Bear Robotics KINO-2: demonstration-to-skill robot-learning platformBear Robotics launched KINO-2 on Oct 21 at GTC Berlin: a robot-learning platform that turns human task demonstrations…robotics · JEV 0.59
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.