Sustainable agriculture is essential to meet the growing food demand. Tomatoes are a key crop for both humans and animals. However, they are vulnerable to infectious diseases. Early diagnosis is crucial to prevent yield losses and maintain food quality. Recent advances in deep learning, especially Convolutional Neural Networks (CNNs), show promise in automating plant disease classification. This research presents a lightweight hybrid convolutional neural network for early tomato leaf disease classification. The proposed model uses multi-scale feature extraction, hybrid feature fusion, and the Swish activation function. It captures fine-grained leaf patterns and broader disease symptoms. The model was evaluated on the PlantVillage Tomato Leaf Dataset. It achieved an accuracy of 98.02%, surpassing VGG-16, ResNet50, and Inception-V3. The model is computationally efficient and lightweight. It can be deployed in mobile and web applications for real-time monitoring. This approach is scalable, cost-effective, and contributes to precision farming and enhancing global food security.