This paper presents a novel deep learning model for the automated classification of brain tumors using magnetic resonance imaging (MRI) data. The model, based on the EfficientNetB3 architecture, was trained and evaluated on a dataset of 7,023 MRI images. The study aims to address the challenges of manual tumor identification, which is time-consuming and prone to errors, by developing an automated system with high accuracy. The preprocessing phase involved dividing the dataset into training, validation, and testing sets, resizing images to 224x244 pixels, normalizing the pixel values, and augmenting the images to improve generalization. The proposed model incorporated batch normalization, a dense layer with regularization, and dropout regularization to prevent overfitting and enhance performance. The results showed that the model achieved 100% training accuracy and 99.6% validation accuracy, indicating its ability to learn key features from the training data. The confusion matrix demonstrated a high number of true positives and a low number of false negatives, along with minimal false positives, confirming the model's high precision and accuracy. Compared to existing methods, the proposed model outperformed current systems, achieving a 100% accuracy rate. This makes it a viable tool for clinical use in categorizing brain tumors based on MRI scans, potentially improving treatment outcomes and reducing the time and effort required for diagnosis. The study highlights the potential of deep learning in medical image analysis, particularly in brain tumor detection, and suggests further research to enhance the model's performance and expand its application in clinical settings.This paper presents a novel deep learning model for the automated classification of brain tumors using magnetic resonance imaging (MRI) data. The model, based on the EfficientNetB3 architecture, was trained and evaluated on a dataset of 7,023 MRI images. The study aims to address the challenges of manual tumor identification, which is time-consuming and prone to errors, by developing an automated system with high accuracy. The preprocessing phase involved dividing the dataset into training, validation, and testing sets, resizing images to 224x244 pixels, normalizing the pixel values, and augmenting the images to improve generalization. The proposed model incorporated batch normalization, a dense layer with regularization, and dropout regularization to prevent overfitting and enhance performance. The results showed that the model achieved 100% training accuracy and 99.6% validation accuracy, indicating its ability to learn key features from the training data. The confusion matrix demonstrated a high number of true positives and a low number of false negatives, along with minimal false positives, confirming the model's high precision and accuracy. Compared to existing methods, the proposed model outperformed current systems, achieving a 100% accuracy rate. This makes it a viable tool for clinical use in categorizing brain tumors based on MRI scans, potentially improving treatment outcomes and reducing the time and effort required for diagnosis. The study highlights the potential of deep learning in medical image analysis, particularly in brain tumor detection, and suggests further research to enhance the model's performance and expand its application in clinical settings.