Brain tumor is one of the most common causes of death in modern times. Early and accurate detection of this disease can save the lives of a large part of the world’s population. Accurate diagnosis and classification of brain tumors in patients using machine learning and deep learning is of great importance in determining effective treatment methods. This article develops the models based on vision transformer, multi-block convolutional neural networks and k-nearest-neighbors, which provide high-precision detection and categorization of brain tumors in patients using magnetic resonance imaging. The main advantage of applying these models is that the processes of feature extraction in images are implemented through attention and filtration mechanisms, rather than the traditional segmentation methods. The proposed models are tested on the Brain Tumor MRI database containing 7023 histological images open for scientific research and evaluated based on various metrics. Comparative analysis of the evaluation results determines a model that identifies all images containing pathological changes with higher accuracy (pp.3-11).