Leukemia, a kind of blood cancer, is difficult to diagnose owing to its many varieties and sophisticated diagnostics. The traditional method of diagnosis is to examine blood and bone marrow samples under a microscope, which takes time. This work investigates the use of deep learning models - VGG16, EfficientNetV2, MobileNetV3, and Densenet121 - to assist in recognizing leukemia cells in blood cell pictures. EfficientNetV2 proved to be the most successful model, with accuracy rates of roughly 99.41% in training, 96.19% in validation, and 96.07% in testing. Integrating these algorithms into diagnostic procedures has the potential to accelerate and increase leukemia detection accuracy. This innovation has the potential to revolutionize leukemia diagnosis, allowing for faster and more accurate detection of malignant cells. Ultimately, it may improve treatment results and patient care by allowing for speedier introduction of suitable medicines. This work represents a major advance in using technology to improve leukemia diagnosis and treatment options.