Building advanced robots at home is no longer a far-fetched dream, thanks to new tools like SmolVLA, an open-source robotics model recently released by AI platform Hugging Face. Designed to be compact yet powerful, SmolVLA brings cutting-edge robotics capabilities to hobbyists and researchers without the need for expensive hardware. Despite having just 450 million parameters, a fraction of the size of larger robotics models, SmolVLA reportedly outperforms many of them in both simulated and real-world environments.
The model is part of Hugging Face’s broader push to make robotics more accessible. Last year, the company launched LeRobot, a growing collection of datasets, tools, and models for robotics development. This effort was further strengthened by the acquisition of France-based Pollen Robotics, a startup known for its open-source hardware. Hugging Face has since introduced several low-cost robotic systems, including humanoids, that developers can experiment with using SmolVLA.
What sets SmolVLA apart is its efficiency and flexibility. It can run on standard consumer GPUs and even on a MacBook, making it one of the most accessible models in its class. The training data for SmolVLA came from community-contributed datasets on Hugging Face’s platform, reflecting the company’s commitment to collaborative development. Importantly, SmolVLA supports asynchronous inference, meaning it can process visual and auditory inputs separately from physical movements. This separation allows robots to react more swiftly in unpredictable environments, a feature that could prove vital in real-world applications.
Already, users are beginning to test its limits. One developer posted on X that they successfully used SmolVLA to control a third-party robotic arm, hinting at the model’s growing appeal among DIY robotics enthusiasts. Hugging Face has made the model publicly available, encouraging experimentation and open collaboration.
While SmolVLA is a breakthrough, Hugging Face is not alone in this space. Tech giants like Nvidia and rising startups such as K Scale Labs, Dyna Robotics, and RLWRLD, along with Jeff Bezos-backed Physical Intelligence, are also racing to shape the future of open robotics. But with its community-driven approach and accessible design, Hugging Face is carving out a unique role in democratising robotics innovation.
