Lung diseases are one of the major causes of increased mortality worldwide, ranging from mild upper respiratory infections to severe respiratory ailments and multi-organ failure. Medical experts face hurdles in diagnosing and managing lung diseases effectively as there are several screening methods including chest X-ray (CXR) examination. Recent technologies aid in efficient lung disease detection, especially using convolutional neural networks (CNN). However, most of those methods can distinguish between normal and one specific lung disease. In this study, we evaluated six deep learning models CheXNet, DenseNet201, DenseNet121, ResNet101, EfficientNetB3, and MobileNetV2 for their ability to classify CXR images into five categories: normal, COVID-19, viral pneumonia, tuberculosis, and lung opacity. Among the models, CheXNet achieved the highest performance, with an AUROC of 0.988 and an accuracy of 94.12%, attributed to its domain-specific pretraining on medical imaging data. EfficientNetB3 followed closely with an AUROC of 0.987 and an accuracy of 93.35%, showcasing its architectural efficiency. MobileNetV2 demonstrated the lowest AUROC of 0.963, highlighting its limitations for complex multi-class classification tasks. These findings emphasize the potential of CNNs, particularly with domain-specific pretraining, for accurate and efficient multi-disease lung diagnostics.