This study evaluates the performance of convolutional neural networks (CNNs) for classifying brain tumors using magnetic resonance imaging (MRI) data. The research compares various CNN architectures, including VGG, ResNet, EfficientNet, MobileNet, and ConvNeXt, on two MRI datasets containing over 3000 images of three tumor types (gliomas, meningiomas, pituitary tumors) and images without tumors. The study assesses the models' accuracy, training time, image throughput, and computational complexity. The results show that several models achieve high accuracy, with the best model achieving 98.7% accuracy. The average precision for each tumor type is 94.3% for gliomas, 93.8% for meningiomas, 97.9% for pituitary tumors, and 95.3% for images without tumors. VGG is the largest model with over 171 million parameters, while MobileNet and EfficientNetB0 are the smallest with 3.2 and 5.9 million parameters, respectively. These two models are also the fastest to train. The study also explores the benefits of transfer learning, data augmentation, and fine-tuning in improving model performance. The results indicate that transfer learning and fine-tuning significantly enhance model accuracy, while data augmentation does not consistently improve accuracy. The study concludes that ResNet, MobileNet, and EfficientNet are the most accurate networks, with MobileNet and EfficientNet demonstrating superior performance in terms of complexity. The findings suggest that CNNs can effectively classify brain tumors using MRI data, with transfer learning and fine-tuning being key factors in achieving high accuracy.This study evaluates the performance of convolutional neural networks (CNNs) for classifying brain tumors using magnetic resonance imaging (MRI) data. The research compares various CNN architectures, including VGG, ResNet, EfficientNet, MobileNet, and ConvNeXt, on two MRI datasets containing over 3000 images of three tumor types (gliomas, meningiomas, pituitary tumors) and images without tumors. The study assesses the models' accuracy, training time, image throughput, and computational complexity. The results show that several models achieve high accuracy, with the best model achieving 98.7% accuracy. The average precision for each tumor type is 94.3% for gliomas, 93.8% for meningiomas, 97.9% for pituitary tumors, and 95.3% for images without tumors. VGG is the largest model with over 171 million parameters, while MobileNet and EfficientNetB0 are the smallest with 3.2 and 5.9 million parameters, respectively. These two models are also the fastest to train. The study also explores the benefits of transfer learning, data augmentation, and fine-tuning in improving model performance. The results indicate that transfer learning and fine-tuning significantly enhance model accuracy, while data augmentation does not consistently improve accuracy. The study concludes that ResNet, MobileNet, and EfficientNet are the most accurate networks, with MobileNet and EfficientNet demonstrating superior performance in terms of complexity. The findings suggest that CNNs can effectively classify brain tumors using MRI data, with transfer learning and fine-tuning being key factors in achieving high accuracy.