This research investigates the application of deep learning models-ResNet101, EfficientNet, ViT, and MobileNet-for the classification of Colorectal Cancer (CRC). CRC is a significant public health concern globally, with early and accurate detection playing a pivotal role in improving patient outcomes. Traditional methods of diagnosis often rely on invasive procedures and histopathological analysis, prompting the exploration of non-invasive, computational approaches. Deep learning models offer promising capabilities in image classification tasks, leveraging their ability to automatically learn discriminative features from medical images. Comparative analyses of ResNet101, EfficientNet, ViT, and MobileNet are conducted to assess their effectiveness in CRC classification tasks. Results indicate varying strengths among the models, highlighting the nuanced trade-offs between computational efficiency and classification accuracy. Ultimately, this research contributes insights into leveraging state-of-the-art transfer learning architectures for enhancing CRC diagnosis, paving the way for future advancements in non-invasive medical imaging and precision oncology.