Blood cancers predominantly impact the generation and operation of blood cells within the organism. Many of these cancers typically initiate in the marrow of bones. Unlike solid tumors seen in other cancers, Acute Lymphoblastic Leukemia (ALL) typically does not manifest as localized tumor growth. ALL detection through medical imaging, particularly using deep learning models, has surfaced as a promising strategy in the realm of cancer diagnosis. Deep learning models have the potential to be put to use medically in order to automate the diagnosis and assist the pathologists in the process. In this paper, different deep learning models including Convolutional Neural Network (CNN), Visual Geometry Group 16 (VGG16), InceptionV3, Inception-ResNetV2, Residual Network 50 (ResNet50), Residual Network 101 (ResNet101), Residual Network 152 (ResNet152), AlexNet, Xception, and Visual Geometry Group 19 (VGG19) are evaluated. Accuracy and F1-score were used for the analysis. The results indicate that the “ResNet152” model demonstrated the best accuracy of 94.25% on the dataset used. By examining multiple families of deep learning architectures, our research provides valuable insights into their performance for cancer diagnosis.