Brain tumors pose life-threatening risks due to their potential to infiltrate neighbouring tissues, emphasizing the critical need for an accurate diagnosis to guide effective treatment plans. Recent advances in deep learning in computer vision have demonstrated significant potential for tumour classification due to the availability of widespread datasets with high-quality annotations. This work explores explicitly the classification of different types of brain tumors using deep learning, focusing on the implementation of EfficientNetB7. The Dataset comprises various Magnetic Resonance Imaging images depicting Glioma, Meningioma, and Pituitary tumors, nearly 7000 images. Notably, the model achieves an impressive overall classification accuracy of 94.43%, outperforming established models such as VGG-16, CNN, Xception, InceptionV3, and DenseNet201. These findings underscore the efficacy of deep learning, particularly with EfficientNetB7, in accurately classifying brain tumors, thereby enhancing diagnostic precision and treatment planning.