Brain tumors poses as a significant global challenge that requires precise diagnostic methods for providing timely treatment decisions. Deep learning models has evolved as powerful tools as it accurately classifies brain tumors in MRI images. The study aims to improve brain tumor classification by identification of following types of tumor—Meningioma, Glioma, Pituitary, and No-Tumor cases using Figshare dataset. This is achieved by following standard practices in data visualization, preprocessing, and dataset partitioning. It uses four advanced transfer learning models—VGG16, ResNet50, EfficientNetB1, and Inception Net for determining the most effective model for the classification task. EfficientNetB1 was found to be the top-performing model by achieving an impressive accuracy of 97.48% closely followed by the VGG16 model with an accuracy of 97.25%. Further architectural modifications were implemented to enhance the EfficientNet B1 model’s accuracy. The resulting model gave an increased accuracy of 99.85%. The proposed model’s exceptional performance underscores its potential utility in medical settings for aiding in accurate diagnosis and treatment planning. Further, continued research efforts will focus on refining model architectures, exploring novel techniques for feature extraction, and further validating the proposed model on larger and more diverse datasets.