Object detection using AI has become inevitable in the present world. YOLO has been used efficiently for object detection since 2015. The present work examines the evolution of the YOLO family of models and conducts object detection comparison of YOLOv8, YOLOv9, and YOLOv10 that have been widely praised for their efficiency and accuracy in application use cases involving real-time. The study investigates enhancements in their architecture, training method, and optimization methods, where every version addresses the limitations of the previous one. Performance comparison on shared benchmarks (Pascal VOC) and prominent metrics, such as computational efficiency, inference time, and mean Average Precision (mAP) has been performed. Besides, model flexibility on various datasets and performance across various deployment use cases—on-edge devices as well as high throughput systems have been investigated with prudence. Findings highlight the evolution of YOLO models, maximizing accuracy, and speed in response to the growing demands of today’s computer vision applications.