The accurate classification of plant diseases is crucial for effective crop management and the advancement of sustainable agricultural practices. Early and precise detection of plant diseases is a cornerstone of precision agriculture, enabling the reduction of crop losses and the optimization of resource allocation. In recent years, deep learning models have shown exceptional performance in image-based tasks, including plant disease classification. However, the integration of advanced attention mechanisms, such as Vision Transformers, offers a promising approach for further enhancing the capabilities of these models. This study proposes a hybrid deep learning approach that combines the strengths of EfficientNet and Vision Transformers to improve the accuracy and efficiency of plant disease classification. This work focuses on using EfficientNet’s ability to extract discriminative local features from leaf images and ViT’s capacity to model long-range dependencies between image patches, thereby refining the classification process. Experiments were conducted on the augmented PlantVillage dataset, comprising 61,486 images of 38 distinct plant diseases across 14 plant species. The proposed hybrid EfficientNet-ViT model achieved a classification accuracy of 93%, with precision, recall, and F1-scores of 91%, 93%, and 92%, respectively, outperforming standalone models such as EfficientNet (89% accuracy) and traditional CNNs (e.g., ResNet: 89%). Comparative analysis with other transformer variants (DeiT, SWIN) further demonstrated the robustness of the approach, with SWIN achieving the highest accuracy (94%). The integration of data augmentation techniques improved model generalizability, contributing to a 4% increase in accuracy over non-augmented training. These results present the potential of combining convolutional neural networks with attention-based mechanisms to address complex challenges in precision agriculture.