Acute Lymphoblastic Leukemia (ALL), a type of blood cancer, can lead to death if not diagnosed early. Traditionally, pathologists use lab results to diagnose ALL by looking through a microscope at the blood sample. However, this approach has certain drawbacks, including being subjective, time-consuming, and prone to mistakes, especially as the workload increases. In contrast, analysis using machine learning techniques (including deep learning) models enhances the detection process and yields precise results. Models based on Convolutional Neural Networks (CNNs) analyze extensive datasets more rapidly and precisely than conventional approaches. They also notify physicians of likely ALL cases, which helps to speed up identification. The proposed research aims to compare the detection accuracy of ALL by utilizing different machine learning models. Four CNN models, including ResNet-34, VGG-16, VGG-19, and ResNet-50, yielded the accuracies 95.07%, 99.22%, 98.18%, and 94.06%, respectively. However, a hybrid model combining ResNet-50 and VGG-16 exceeded all state-of-the-art frameworks with the highest accuracy of 99.37%.