Hugging Face Claims New Robot Model Is Efficient Enough to Run on a MacBook

In recent developments, creating complex robotics projects at home has become somewhat more accessible.

Earlier this week, AI development platform Hugging Face unveiled SmolVLA, an open AI robotics model. According to Hugging Face, SmolVLA outperforms larger robotics models in both virtual and real-world environments. The model is trained on a community-shared dataset with a "compatible license."

"SmolVLA aims to make vision-language-action (VLA) models more accessible and accelerate research into general-purpose robotic agents," Hugging Face wrote in their blog post. "SmolVLA is not only lightweight yet powerful, but also offers a way to train and evaluate general robotics technologies."

SmolVLA is part of Hugging Face's rapidly growing ecosystem of affordable robotics hardware and software. Last year, the company launched LeRobot, a collection of robotics-focused models, datasets, and tools. Recently, Hugging Face acquired Pollen Robotics, a French robotics startup, and introduced several low-cost robotic systems, including humanoid robots, available for purchase.

With 450 million parameters, SmolVLA was trained on datasets from the LeRobot community—specially labeled robotics datasets from Hugging Face's AI development platform. Parameters, sometimes referred to as "weights," are internal components that guide a model’s behavior.

Hugging Face claims that SmolVLA is compact enough to run on a single consumer-grade GPU or even a MacBook, allowing it to be tested and deployed on "affordable" hardware, including the company’s robotic systems.

Interestingly, SmolVLA also supports an "asynchronous inference stack," which Hugging Face says separates the processing of a robot's actions from what it sees and hears. As the company explained in its blog, "This separation allows robots to respond faster in rapidly changing environments."

SmolVLA is available for download on Hugging Face.

Notably, Hugging Face is far from the only player in the emerging open robotics field.

Nvidia offers a suite of open robotics tools, while startup K-Scale Labs is developing components for what they call an "open-source humanoid robot." Other key players in this space include Dyna Robotics, Physical Intelligence (backed by Jeff Bezos), and RLWRLD.