Diabetic foot ulcers (DFUs) represent a critical complication of diabetes, often leading to severe outcomes such as infections, amputations, and increased mortality if not properly managed. Accurate and timely grading of DFUs is essential for effective treatment and intervention. In this study, we introduce GradTX Net, a Vision Transformer-based framework designed for categorizing the severity of diabetic foot ulcers into five distinct classes: Grade 0, Grade 1, Grade 2, Grade 3, and normal class. The method involves segmenting DFU images into non-overlapping patches, which are subsequently flattened and transformed into sequences. These sequences are then subjected to linear and positional embedding processes to retain spatial information. The embedded sequences are processed through multiple multi-head attention layers to derive the final feature representation. During the classification phase, the first token in the sequence is passed through a SoftMax layer to generate the classification output. Our experiments, conducted on a modest dataset of DFU images, reveal that GradTX Net achieves superior performance compared to traditional convolutional neural network (CNN) methods, with an accuracy of 95%. These findings underscore the promise of transformer-based architectures in the medical domain, particularly for the automated grading of DFUs.