YOLOv10 is a next-generation object detection model that achieves high accuracy while maintaining real-time performance. Through optimized post-processing and model architecture, it reduces computational redundancy, improving efficiency and performance. YOLOv10 achieves state-of-the-art performance and efficiency across different model scales. For example, YOLOv10-S achieves 1.8x speed improvement compared to RT-DETR-R18 at similar AP, while reducing the number of parameters and FLOPs by 2.8x.