In medical imaging, classification is very crucial and majorly create impact on the diagnosis specifically in brain tumor. Convolutional model is considering as the base model in deep learning, proposed architecture follows hybrid strategy which combines Vision Transformer (ViT) and Convolutional Neural Networks (CNNs) to capitalize on the complementing advantages of both architectures. To investigate their potential for effectively diagnosing brain tumors, six hybrid models were created: CNN + LSTM, LSTM + ViT, EfficientNet + ViT, ViT + Auto Encoder, Auto Encoder + EfficientNet, and CNN + ViT. The approach successfully identifies both local and global patterns in medical images by fusing the self-attention mechanisms of ViT with the feature extraction powers of CNN based on the analysis 97.33% accuracy achieved which is used to classify tumor image precisely. This work underscores the potential of hybrid deep learning architectures to advance brain tumor diagnosis, offering a robust and scalable framework for real-world medical imaging applications.