Early and correct identification of plant diseases is highly vital for sustainable agriculture, especially in highvalue crops such as mangoes. This research presents a Hybrid InceptionV3 plus Vision Transformer (ViT) model to automate mango leaf disease classification with high accuracy. Capitalizing on a large dataset of high-resolution images of the Rajbari and Pabna districts in Bangladesh, the model categorizes mango leaves into six disease classes: Anthracnose, Dried, Gall Midge, Healthy, Powdery Mildew, and Senescent. The hybrid model outperforms standalone models like MobileNetV2, EfficientNetV2B0, and InceptionV3 through the combination of local feature extraction abilities of InceptionV3 with global context sensitivity of ViT, rendering a whopping 98.89% accuracy. The model's effectiveness is also proven by ROC curve analysis with AUC values getting near 1.0000 for all but one class. The approach shows the power of merging CNN and transformer models, providing a robust solution to machine learning-based detection of mango leaf disease, which could have a significant effect on crop management and help advance precision agriculture.