Automatic License Plate Recognition (ALPR) is crucial for modern Intelligent Transportation Systems (ITS), but previous models like YOLOv8 face limitations in efficiency for real-time deployment on edge devices. To address this, our study proposes an ALPR system combining the advanced YOLOv12 model with EasyOCR, validated on a custom dataset of 8,397 diverse license plate images. Comparative analysis demonstrates YOLOv12’s superiority, with its lightweight Yolov12n variant achieving a higher mAP50–95 score (0.756) than Yolov8n (0.746), indicating enhanced localization accuracy. This study’s key contribution is providing a benchmark that confirms YOLOv12’s architectural advancements are not only for accuracy but are pivotal for enabling high-performance, real-time ALPR deployment. The findings establish YOLOv12 as a more effective and efficient solution for resource-constrained systems.