Brain tumors are life-threatening conditions that require early and accurate detection for effective treatment planning. Traditional diagnostic methods, such as MRI scans and biopsies, depend heavily on radiologist interpretation, making the process time-consuming and susceptible to errors. Deep learning has revolutionized medical imaging by automating tumor detection and classification. Although Convolutional Neural Networks (CNNs) have been widely used, their limited capacity to capture long-range dependencies in images can hinder performance. Vision Transformers (ViT-B16) offer a promising alternative by processing images as sequences of patches, allowing for better analysis of global contextual information. This paper presents a deep learning-based framework for brain tumor classification using ViT-B16. The model leverages transfer learning and pretraining on large-scale datasets to enhance feature extraction and improve classification accuracy. A comparative analysis with CNN-based models, including VGG-16, ResNet-50, and YOLOV10, shows that ViT-B16 achieves superior accuracy and reliability in brain tumor detection. This study contributes to improving early tumor diagnosis, streamlining the medical imaging workflow, and advancing AI-driven healthcare solutions.