This paper presents a hybrid deep CNN model for multi-classification of brain tumors. The study aims to develop an automated, deep learning-based system for classifying brain tumors, which is more accurate and efficient than traditional methods. Three distinct CNN models are proposed for different classification tasks. The first model achieves 99.53% detection accuracy for brain tumors. The second model, with 93.81% accuracy, classifies brain tumors into five types: normal, glioma, meningioma, pituitary, and metastatic. The third model achieves 98.56% accuracy in classifying brain tumors into their different grades. Grid search optimization is used to automatically fine-tune the hyperparameters of the CNN models. The models are trained on large, publicly available clinical datasets, resulting in robust and reliable classification outcomes. The study compares the proposed models with classical models such as AlexNet, DenseNet121, ResNet-101, VGG-19, and GoogleNet, reaffirming the superiority of the deep CNN-based approach in brain tumor classification and early detection. The models are evaluated using various performance metrics, including accuracy, sensitivity, precision, and the area under the ROC curve. The results show that the proposed CNN models outperform other models in all classification tasks. The study concludes that the proposed CNN models are effective in classifying brain tumors and can be used to assist clinicians and radiologists in early detection and diagnosis.This paper presents a hybrid deep CNN model for multi-classification of brain tumors. The study aims to develop an automated, deep learning-based system for classifying brain tumors, which is more accurate and efficient than traditional methods. Three distinct CNN models are proposed for different classification tasks. The first model achieves 99.53% detection accuracy for brain tumors. The second model, with 93.81% accuracy, classifies brain tumors into five types: normal, glioma, meningioma, pituitary, and metastatic. The third model achieves 98.56% accuracy in classifying brain tumors into their different grades. Grid search optimization is used to automatically fine-tune the hyperparameters of the CNN models. The models are trained on large, publicly available clinical datasets, resulting in robust and reliable classification outcomes. The study compares the proposed models with classical models such as AlexNet, DenseNet121, ResNet-101, VGG-19, and GoogleNet, reaffirming the superiority of the deep CNN-based approach in brain tumor classification and early detection. The models are evaluated using various performance metrics, including accuracy, sensitivity, precision, and the area under the ROC curve. The results show that the proposed CNN models outperform other models in all classification tasks. The study concludes that the proposed CNN models are effective in classifying brain tumors and can be used to assist clinicians and radiologists in early detection and diagnosis.