This study evaluates the performance of a real-time detection transformer (RT-DETR) object detection algorithm in identifying *Plasmodium* species on thin blood smear images. The algorithm was trained and validated on a dataset of 24,720 images from 475 thin blood smears, and tested on a dataset of 4,508 images from 170 smears. At the patient level, the RT-DETR algorithm achieved an overall accuracy of 79.4% with a recall of 74% for negative smears and 81.9% for positive smears. Among *Plasmodium*-positive smears, the global accuracy was 82.7% with a recall of 90% for *P. falciparum*, 81.8% for *P. malariae*, and 76.1% for *P. ovale/vivax*. The RT-DETR model met the World Health Organization (WHO) competence level 2 for species identification and can run in real-time on low-cost devices like smartphones, making it suitable for deployment in low-resource settings. The study also compared the RT-DETR algorithm with YOLOv5 and YOLOv8, finding that RT-DETR performed similarly at the patient level but had slightly better label-level performance. The results suggest that the RT-DETR algorithm could be a valuable tool for improving malaria diagnosis, especially in areas with limited access to skilled microscopists.This study evaluates the performance of a real-time detection transformer (RT-DETR) object detection algorithm in identifying *Plasmodium* species on thin blood smear images. The algorithm was trained and validated on a dataset of 24,720 images from 475 thin blood smears, and tested on a dataset of 4,508 images from 170 smears. At the patient level, the RT-DETR algorithm achieved an overall accuracy of 79.4% with a recall of 74% for negative smears and 81.9% for positive smears. Among *Plasmodium*-positive smears, the global accuracy was 82.7% with a recall of 90% for *P. falciparum*, 81.8% for *P. malariae*, and 76.1% for *P. ovale/vivax*. The RT-DETR model met the World Health Organization (WHO) competence level 2 for species identification and can run in real-time on low-cost devices like smartphones, making it suitable for deployment in low-resource settings. The study also compared the RT-DETR algorithm with YOLOv5 and YOLOv8, finding that RT-DETR performed similarly at the patient level but had slightly better label-level performance. The results suggest that the RT-DETR algorithm could be a valuable tool for improving malaria diagnosis, especially in areas with limited access to skilled microscopists.