With the rapid development of the electronics manufacturing industry, defect detection for printed circuit boards (PCBs) is crucial for product quality control. Traditional inspection methods rely on manual visual inspection or traditional image processing techniques, which suffer from low efficiency and high missed detection rates. Although deep learning-based object detection algorithms have significantly improved detection accuracy, existing models suffer from large parameter counts and high computational complexity, making them difficult to meet the real-time and lightweight deployment requirements of industrial scenarios. To address this, this paper proposes a lightweight PCB defect detection algorithm based on YOLOv11, the YoloV11-CGL model. First, the Context-Guided module is introduced to replace the existing C3k2 module. This module utilizes a multi-stage context fusion mechanism and efficient channel separation computation to significantly reduce parameter size while maintaining pixel-level classification accuracy. Second, the LAE module performs down-sampling, replacing the existing down-sampling. This module utilizes adaptive weight fusion and channel information reorganization to dynamically retain high-entropy pixels during the four-fold down-sampling process, avoiding the edge feature loss of traditional convolution and reducing the number of parameters to 1/N of the traditional method (N is the number of group convolutions). Furthermore, to address the accuracy degradation caused by lightweight models, knowledge distillation is employed to effectively improve model accuracy while minimizing the number of parameters. Experimental results demonstrate that the YoloV11-CGL model achieves excellent results on the public DsPCBSD+ datasets. While significantly reducing the number of parameters by 26%, accuracy remains largely unchanged, making it suitable for the real-time and lightweight deployment requirements of industrial scenarios.