Abstract: Wafer defect detection is a crucial step in semiconductor manufacturing, directly affecting product yield and production efficiency. Existing wafer defect detection algorithms generally have difficulties in balancing detection accuracy, lightweight design, and inference speed. Moreover, they often suffer from missed detections and incorrect judgments in mixed defect scenarios. Based on this, to meet the requirements of wafer mixed defect detection, using the YOLOv11n model as the core, performance verification experiments were conducted on single and mixed defect detection of wafers. Firstly, addressing issues such as imbalance in defect categories and missing labels in the open-source dataset, data screening and data augmentation were carried out on the MixedWM38 dataset, resulting in the construction of a wafer defect detection dataset containing 15 700 samples. Secondly, single defect average detection accuracy was quantified using mAP@0.5 and mAP@0.5:0.95, while mixed defect detection accuracy was quantified using Hamming loss, precision match rate, etc. Parameters and GFLOPs were used to evaluate model lightweighting, and FPS was used to represent model inference speed. Finally, in the same experimental conditions, the YOLOv11n model was compared with typical object detection models such as SSD, Faster R-CNN, RT-DETR-L, and YOLOv8n. The experimental results show that YOLOv11n demonstrates significant advantages in all evaluation indicators and achieves the optimal balance in detection accuracy, lightweight design, and inference speed. mAP@0.5, mAP@0.5:0.95, Parameters, GFLOPs, and FPS reached 99.4%, 92.4%, 2.6M, 6.3, and 99.3, respectively. This research provides a reliable technical solution for high-precision real-time detection of wafer mixed defects in semiconductor industrial production lines.