Early detection and accurate diagnosis of plant diseases are crucial for ensuring food security and maximizing agricultural productivity. Over the last couple of years, deep learning techniques have been increasingly promising at automating plant disease diagnosis. This paper explores how four cutting-edge deep learning models, including VGG16, Recurrent Neural Networks (RNN), Vision Transformers (ViTs), and Capsule Networks (CapsNet), when applied to a dataset of 87,000 RGB images for 38 different classes are able to be compared. The performance of all models is assessed using accuracy, precision, recall, F1-score, and computational efficiency. We also use an AI-powered chatbot that helps farmers and agricultural experts in diagnosing plant diseases through inputs such as images and text. The experimental results show which model is better and which is weaker, and Vision Transformers and Capsule Networks feature extraction is better than traditional CNNs. By integrating the chatbot, the system becomes more accessible and much easier for real-world applications.