Deep learning algorithms are essential for rapid and reliable diagnosis of brain tumors (BT), a crucial task in medical imaging. Here, we offer a comparison method to investigate how well Vision Transformers and Convolutional-Neural-Networks (ViCNNs) perform in improving the detection of brain tumors. Our research shows encouraging outcomes, suggesting substantial improvements in model performance and classification accuracy. We proposed a DL model ViTat 1.0 enriched with attention-based Vision Transformer (ViT) mechanism to detect medical images’ global dependencies. In addition to this, we also tested ViTat 1.0 and other studied CNN models on Generative Adverbial Networks (GANs) augmented images and observed a significantly increase (more than 5%) in classification accuracy. The results show Consistent and higher performance, which confirms that the proposed ViTat 1.0 is practical across different datasets. Results show that ViTat 1.0 exceeds CNNs, proving once again how critical it is to use cutting-edge DL architectures for medical image processing. Finally, our study enhances the field of BT identification by demonstrating the effectiveness of GAN and attention with ViT in improving classification accuracy and model performance. With these results, ViTat 1.0 can be seen as an application that could improve medical image processing and lead to faster, more precise clinical diagnosis.