Diabetic retinopathy is an eye disease related to retinal damages because of high blood sugar levels. This disease is not curable directly, but an early detection can help to protect the vision of the sufferer. Deep learning models using image processing technologies over the fundus images of the eye are used to detect this condition. This study proposes a ViT (Vision Transformer) based deep model for retinopathy detection as a two-class classification problem. Aptos 2019 dataset is used to build the model, and the different severity levels of retinopathy are treated as presence of disease as a single class. The proposed model is assessed for its accuracy, F-measure and recall/precision. The area under the curve. The model utilizes data preprocessing and augmentation to improve the quality of datasets. The results demonstrate that proposed ViT based deep model for diabetic retinopathy detection with accuracy of 98.0% and AUC of 99.5%. It is concluded that deep learning models based on vision transformers have significant potential to detect diabetic retinopathy to help the patients to save their vision and lead a better life.