SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Robotics

Benchmark Model Rank Results
robot-manipulation-on-libero-objectSmolVLA (0.45B)#89Success Rate: 96
robot-manipulation-on-libero-objectSmolVLA (2.25B)#95Success Rate: 94
robot-manipulation-on-libero-spatialSmolVLA (2.25B)#89Success Rate: 93
robot-manipulation-on-libero-spatialSmolVLA (0.45B)#96Success Rate: 90
robot-manipulation-on-libero-goalSmolVLA (0.45B)#90Success Rate: 92
robot-manipulation-on-libero-goalSmolVLA (2.25B)#94Success Rate: 91
robot-manipulation-on-libero-longSmolVLA (2.25B)#97Success Rate: 77
robot-manipulation-on-libero-longSmolVLA (0.45B)#101Success Rate: 71
robot-manipulation-on-liberoSmolVLA (2.25B)#126Success Rate: 88.75