• Hugging Face unveils SmolVLA, an open-source Vision-Language-Action (VLA) model designed to run on affordable hardware like MacBooks and consumer GPUs, making robotics development more accessible.
  • Trained on 23,000 episodes from 481 datasets, SmolVLA integrates vision, language, and action capabilities, enabling robots to interpret instructions, perceive environments, and perform physical tasks efficiently.

Hugging Face, an AI development platform, has introduced SmolVLA, a compact, open-source Vision-Language-Action (VLA) model for cost-effective and efficient robotics, in a groundbreaking effort to democratize robotics.

By operating on inexpensive hardware, such as consumer-grade GPUs or even MacBooks, this novel concept, unveiled on June 5, 2025, aims to make sophisticated robotic systems accessible to developers, academics, and enthusiasts.

“SmolVLA aims to democratize access to vision-language-action [VLA] models and accelerate research toward generalist robotic agents,” writes Hugging Face. “SmolVLA is not only a lightweight yet capable model, but also a method for training and evaluating generalist robotics .”

With 450 million parameters, SmolVLA was trained on about 23,000 episodes from 481 community datasets from Hugging Face's LeRobot platform, which is a suite of models, datasets, and tools centered around robotics that was unveiled last year.

By combining vision, language, and action skills, the approach allows robots to process text instructions, analyze visual inputs, and carry out physical activities.

It is a workable solution for real-world applications because of its small size and innovative optimization techniques, which guarantee quicker performance and reduced latency.

Hugging Face's move into robotics expands upon its competence in humanoid robot design, which was strengthened by its April 2025 acquisition of Pollen Robotics.

According to the company, SmolVLA will power reasonably priced robotic systems like the recently released HopeJR and Reachy Mini, which retail for about $3,000 and $250–$300, respectively.

With their modular designs and completely configurable features, these open-source robots are meant to compete with more expensive rivals such as Tesla's Optimus Gen 2 and Unitree's G1.

Nvidia has a set of open robotics tools, and K-Scale Labs, a startup, is constructing the parts for what it refers to as "open-source humanoids." Other strong companies in the market include RLWRLD, Jeff Bezos-backed Physical Intelligence, and Dyna Robotics.

This launch represents a major advancement in accessible, reasonably priced robots that could revolutionize home applications, education, and industry.

Edited by Harshajit Sarmah