Colon cancer must be accurately and promptly detected in order to improve patient outcomes and direct the development of effective treatment plans. In this paper, a unique Modified Vision Transformer (MViT) architecture designed for automated colon histopathology image classification into benign and cancer categories is presented. The MViT efficiently extracts global contextual information from intricate tissue structures by incorporating unique improvements into the conventional transformer design A publicly accessible dataset was used to train and evaluate the model, which produced an astounding 99.3% classification accuracy. It continuously outperformed a number of leading-edge deep learning and hybrid models in important assessment measures, such as ROC-AUC (0.99), precision (0.9940), recall (0.9920), and F1-score (0.9930). These outcomes highlight the model's strong discriminative power and dependability. All things considered, the suggested MViT framework offers a strong and effective method for analysing histopathological images and has a great chance of being integrated into computer-aided diagnostic (CAD) systems to improve clinical judgment when diagnosing colon cancer.