SmolVLA: Efficient Vision-Language-Action Model trained on Lerobot Community Data
SmolVLA is a compact (450M) open-source Vision-Language-Action model for robotics, trained on community data, that outperforms larger models on simulation and real-world tasks.
