Creating sophisticated robotics projects at home just became significantly easier thanks to the newest development from AI platform Hugging Face. In a notable advancement for robotics, Hugging Face has released a novel, open-source artificial intelligence model named SmolVLA, designed specifically for robotics applications. According to the company, the compact new model surpasses much larger robotics models previously available, performing exceptionally in both simulated and real-world scenarios.
SmolVLA, a “vision-language-action” (VLA) model comprising 450 million parameters, was trained using publicly available and compatibly licensed datasets shared by Hugging Face’s growing LeRobot Community. Model parameters, sometimes called “weights,” serve as the decision-making components which drive the model’s actions and behaviors.
Hugging Face emphasizes SmolVLA’s efficiency, touting its ability to run smoothly on consumer-grade hardware such as standard GPUs and even laptops like a MacBook. This capability removes many previous technical barriers, enabling a broader range of users to experiment with, test, and deploy sophisticated robotics technologies affordably and at scale.
A unique technical feature of SmolVLA is its “asynchronous inference stack,” which enables robotic systems to independently handle decision-making processes and sensory analyses. According to Hugging Face, this architectural refinement allows robots to respond swiftly in rapidly evolving environments, providing heightened efficiency and flexibility.
SmolVLA is part of a broader strategic push by Hugging Face to democratize access to robotics technology, an initiative that has gathered significant momentum over the past year. The company’s LeRobot suite already comprises robotics-focused models, datasets, and developer tools. Recently, Hugging Face expanded its robotics footprint through the acquisition of the French startup Pollen Robotics and releasing affordable robotics hardware, including humanoid models and low-cost 3D-printed robotic arms.
Reaction to SmolVLA has been swift and positive within the robotics community. A developer has already shared online their success integrating SmolVLA with a third-party robotic arm using modest hardware. Hugging Face’s growing activity places the company squarely within a rapidly evolving ecosystem of open-source robotics initiatives, competing alongside major industry players such as Nvidia, Dyna Robotics, Jeff Bezos-funded startup Physical Intelligence, RLWRLD, and K-Scale Labs, which has positioned itself as a proponent of open-source humanoid robotics.
SmolVLA is accessible immediately through Hugging Face’s platform, marking another significant step toward accessible, capable robotics research and development for enthusiasts, researchers, developers, and innovators around the world.
