Detecting and diagnosing cancer cells especially in brain tumor is a very complex process, but it is an effective selection of treatment. Malignant tumors can metastasize, while benign tumors are confined and do not spread. This paper proposes a new method of classifying brain tumors with the advanced neural net model namely EffcientNet-B0 and ResNet-50. The classification system divides tumors into four types, namely specific meningioma tumor, pituitary tumor, glioma tumor, and no tumor. These models were trained on broader image datasets, which are employed to examine Magnetic Resonance Imaging (MRI) images. The pre-trained EfficientNet-B0 and ResNet-50 were used as methods for basic comparison with other networks and achieved an accuracy up to 98%.