Runway introduces Praxis-1 world action model for robotics

Runway AI Inc. introduced Praxis-1, an open-weight world action model that turns the company's large-scale video pretraining into robot control, and plans to release it publicly with open weights in the coming months. Runway chief technology officer Kamil Sindi said Praxis-1 learns mostly from third-person video rather than scarce robot demonstrations, and the company reported that simulating robot policies inside its world model predicts real-world results with 0.95 correlation. Runway has been testing Praxis-1 with early partners, including Noble Machines, ahead of the public launch.

Runway introduces Praxis-1 world action model for robotics The Robot Report https://www.therobotreport.com When Runway does release Praxis-1 publicly, it plans to ship it with open weights rather than as a closed model. The post Runway introduces Praxis-1 world action model for robotics appeared first on The Robot Report. Runway AI Inc. this week introduced Praxis-1, an open- weight https://www.machinebrief.com/glossary/weight world action model that turns Runway’s video pretraining into control for robots. The company said it built Praxis-1 on the same large-scale video pretraining behind its general world models. “Most robot policies are bottlenecked by robot data, which is scarce and expensive to collect,” Kamil Sindi, Runway’s chief technology officer, told The Robot Report https://www.therobotreport.com/ . “Praxis-1 learns mostly from third-person video, built on the same large-scale pretraining behind Runway’s world models, so it already understands how objects behave and how tasks unfold. “Also, performance improves as we scale video, so its ceiling is set by how much video it can learn from, not how many robot demonstrations exist,” he added. Founded in 2018, Runway AI specializes in generative artificial intelligence https://www.machinebrief.com/glossary/artificial-intelligence research and technologies. The company https://runway.com/ offers a range of AI https://www.therobotreport.com/category/design-development/ai-cognition/ models, including Aleph 2.0 for in-context video editing, Act-Two for motion capture, and Gen-4.5 for text-to-video and image-to-video generation. For robotics https://www.machinebrief.com/category/robotics , Runway also offers GWM-1 https://runway.com/research/introducing-runway-gwm-1 , a general world model https://www.machinebrief.com/glossary/world-model . The Brooklyn, N.Y.-based company has offices in New York, San Francisco, Seattle, London, Paris, Tel Aviv, and Tokyo. Runway has been testing Praxis-1 with early partners and plans to release the model publicly in the coming months. Editor’s note: Physical AI is among the session topic tracks at RoboBusiness https://www.robobusiness.com/ 2026, which will be on Oct. 20 and 21 in Santa Clara, Calif. Register now to attend. https://cvent.me/5AGyXE?RefId=articles Runway bets on video data for robot training https://www.machinebrief.com/glossary/training While there is a lack of real-world robotics data to train generalist AI models, there is an abundance of video data. Every day, people film and upload more of everyday life each day than any robot lab could capture through teleoperated demonstrations, noted Runway AI. Runway has found that simulating https://www.therobotreport.com/category/software-simulation/ robot policies inside its world model predicts real-world results with 0.95 correlation. This compares favorably with more expensive 3D reconstruction–based techniques, the company claimed. In addition, Runway has extended its work on pretraining large video models into interactive, real-time video models like Solaris and GWM Worlds 2. By teaching models how to generate accurate physics, how hands move, what a task looks like partway through, the company said it has created dynamic, complex environments for agent training in the digital and physical world. Runway said Praxis-1 brings the same approach to robotics, providing a generalist policy model for robotics developers and researchers that works across any embodiment or environment. “Praxis-1 has been trained on a variety of manipulation tasks, ranging from straightforward pick-and-place actions, such as lifting soda cans, to more complex tasks involving deformable objects, like packing gift bags,” Sindi said. Praxis-1 is already testing with early partners Runway is rolling Praxis-1 out to key partners ahead of a public launch. The company is also hoping to provide access to additional partners pre-launch. Runway plans to evaluate its model across a variety of robots. “Early partners, including Noble Machines https://www.therobotreport.com/tag/noble-machines/ , Standard Bots https://www.therobotreport.com/tag/standard-bots/ , and Ultra, are running Praxis-1 on their own hardware,” Sindi said. “A big takeaway so far is that a single model can adapt across very different embodiments, from bimanual arms to humanoids, with light fine-tuning https://www.machinebrief.com/glossary/fine-tuning . Their testing is helping us identify and close gaps before general availability.” When Runway does release Praxis-1 publicly, it plans to ship it with open weights rather than as a closed model. “We believe U.S. leadership in physical AI is critical to regaining our manufacturing lead, and that requires open American models,” Sindi said. “Open weights give hardware developers flexibility and control they don’t have today.” Looking ahead, the company hopes to pre-launch with more select partners to continue improving its model ahead of its full launch. “We’ll continue testing and evaluating Praxis-1 with partners, bring on more early-access partners, and assess efficacy and safety across different embodiments and environments,” Sindi said. The post Runway introduces Praxis-1 world action model for robotics https://www.therobotreport.com/runway-introduces-praxis-1-world-action-model-robotics/ appeared first on The Robot Report https://www.therobotreport.com . Get AI news in your inbox Daily digest of what matters in AI. Key Terms Explained Artificial Intelligence https://www.machinebrief.com/glossary/artificial-intelligence The science of creating machines that can perform tasks requiring human-like intelligence — reasoning, learning, perception, language understanding, and decision-making. Fine-Tuning https://www.machinebrief.com/glossary/fine-tuning The process of taking a pre-trained model and continuing to train it on a smaller, specific dataset to adapt it for a particular task or domain. Training https://www.machinebrief.com/glossary/training The process of teaching an AI model by exposing it to data and adjusting its parameters to minimize errors. Weight https://www.machinebrief.com/glossary/weight A numerical value in a neural network that determines the strength of the connection between neurons.