1 paper
Ailin Huang, Chengyuan Yao, Chunrui Han +90
We present STEP3-VL-10B, a lightweight open-source foundation model designed to redefine the trade-off between compact efficiency and frontier-level multimodal intelligence. STEP3-…