Accurate brain tumor classification is crucial for precise diagnosis and effective treatment planning. Despite the extensive use of comparative analyses of deep learning models in the literature, no prior studies have utilized data from the National Guard Hospital, which provides a unique and clinically relevant dataset. This research aims to address this gap by evaluating the performance of various deep learning models on this specific dataset to identify the most accurate and reliable model for brain tumor classification. This study leverages transfer learning techniques in Convolutional Neural Networks (CNNs), employing pre-trained models to enhance the classification process. By fine-tuning models such as VGG16, ResNet50, InceptionV3, and Xception, the research capitalizes on the strengths of these architectures trained on large-scale datasets to improve classification accuracy in a novel, real-world setting. The comparative analysis conducted in this research highlights the performance of these CNN models on the National Guard Hospital’s dataset, filling a critical gap in the literature where this specific dataset has been previously underexplored. Results show that VGG16 achieves the highest accuracy at 91.47%, followed by ResNet50 at 86.80%, InceptionV3 at 82.67%, and Xception at 82.13%. These findings demonstrate the potential of transfer learning models in enhancing brain tumor classification accuracy when applied to specialized clinical data, ultimately contributing to improved patient care.