This study explores the use of Vision Transformers (ViTs) for breast cancer classification, leveraging the BreakHis dataset containing histopathological images. ViTs, known for their self-attention mechanisms, were fine-tuned using pre-trained models to enhance their ability to analyze and classify breast cancer images across varying magnification levels. The results demonstrate that ViTs significantly outperform traditional convolutional neural network (CNN)-based models in terms of both classification accuracy and robustness to changes in image resolution and scale. Comparative experiments highlighted the superiority of ViTs in capturing long-range dependencies and complex patterns within histopathological data, addressing limitations often encountered with CNN architectures. Our experimental results achieved a peak validation accuracy of 96%, underscoring the potential of ViTs to transform breast cancer diagnosis. Furthermore, the study provides insights into optimizing ViT architectures for medical applications, paving the way for their integration into automated diagnostic systems. This research highlights the promise of ViTs as a robust, high-accuracy tool for advancing breast cancer classification and improving clinical decision-making in pathology.