Brain tumor classification is crucial for accurate diagnosis and treatment planning, and this study investigates the application of an ensemble of deep learning models to improve its reliability in MRI images. The research addresses the challenge of accurately classifying brain tumors and aims to enhance diagnostic precision through advanced AI techniques. The study trained and evaluated six pre-trained convolutional neural network models: VGG16, MobileNet, DenseNet121, DenseNet201, InceptionV3, and ResNet50—using a dataset that comprised four tumor classes: glioma, meningioma, pituitary tumor, and no tumor. The ensemble model was constructed from the top three performing models, achieving significant improvements in classification accuracy (97.35% accuracy) and precision over individual models. This approach highlights the viability of ensemble learning in medical imaging, presenting a robust diagnostic tool with the potential to significantly enhance AI-assisted healthcare and clinical diagnostics. By offering a more accurate method for brain tumor detection, this study contributes valuable insights into the future of medical imaging and diagnosis, paving the way for further research in improving diagnostic accuracy and patient outcomes. Future work will explore expanding the dataset and integrating other ensemble learning methods to further enhance diagnostic performance and generalizability across diverse medical imaging data.