SmolVLA: Efficient Vision-Language-Action Model trained on Lerobot Community Data

Hugging Face Blog · 1 year ago

SmolVLA is a 450-million-parameter open-source vision-language-action model for robotics that runs on consumer hardware and uses only publicly available datasets. The model was trained on fewer than 30,000 episodes—roughly one-tenth the data of comparable systems—yet matches or exceeds the performance of much larger models on simulation and real-world robotics tasks. The asynchronous inference system enables 30% faster response times and 2× task throughput by decoupling action execution from perception processing, allowing robots to respond more quickly to changing environments.