MRI image classification has seen significant advancements using convolutional neural networks (CNNs), particularly in realm of Alzheimer’s Disease and brain tumor identification. The research presents a novel method to enhance MRI image classification performance by leveraging EfficientNetB3, a CNN architecture known for its balanced combination of accuracy and computational efficiency, EfficientNetB3 has been utilized for the categorization of MRI images associated with brain illnesses. This approach enhances the search and resource allocation, without depending on traditional optimization techniques. The suggested model underwent assessment using two datasets: the Kaggle Alzheimer’s Disease International dataset and a separate database for brain tumors. The experimental findings indicate that our model attains a 99.00% accuracy for both the AD and Brain Tumor dataset, respectively. These outcomes showcase the effectiveness of EfficientNetB3 in enhancing the precision and accuracy of MRI image analysis for identification of brain abnormalities.