Accurately detecting objects is vital to success in both safety and the reliability of autonomous driving systems. To explore this need, YOLOv8, YOLOv11, and YOLOv12 were compared using the KITTI dataset, which captures a range of driving scenarios and multiple object classes. All models were trained and tested under the same experimental conditions for comparison. Standard performance metrics, including precision, recall, and mean Average Precision (mAP) at different intersection over Union IoU levels, were analyzed, while also controlling for mAP@0.5.The benchmarking covered the process of pre-processing the KITTI dataset into a structure for the YOLO model training process, while ensuring that the models were trained with the same hyperparameters. The results were evaluated with the validation dataset by measuring all detections made by the models in the vehicles. The comparative analysis highlighted the trade-offs between accuracy and reliability for each detector. Overall, the results indicated that YOLOv12 had demonstrated overall accuracy detection with an mAP@0.5 score of 90.70%, YOLOv11 had the highest recall of the three models at 84.02%, and YOLOv8 produced the highest precision score of the 3 classes at 92.58%. Generally, these results indicated that none of the models had produced any dominant measure for the assessment in the metrics discussed, and the choice of model will come down to those key measures that are suited for each application, and whichever model is assessed for precision.