Timely treatment and improved patient outcomes depend on the accurate and quick diagnosis of lung disorders from chest X-ray images. However, radiologists may need a lot of effort and uneven interpretation to manually identify and interpret these images. In the analysis of medical images, deep learning techniques—particularly transfer learning approaches—have revealed encouraging outcomes. Inception v3, ResNet101v2, DenseNet121, VGG19, and MobileNet are the five notable transfer learning models whose performance is evaluated with Nadam optimizer in this study’s classification of lung diseases into four groups: COVID-19 positive, normal, viral pneumonia, and lung opacity. To train and assess the models, we employed a dataset of chest X-ray images from Kaggle repository. We fine-tuned the pre-trained models on this dataset and assessed their performance using various metrics, such as accuracy, loss, precision, recall, F1-score, and area under the receiver operating characteristic curve (AUC). The outcomes show MobileNet provides best performance of all the five models with classification accuracy, AUC, precision, recall, and F1-score of 90.39%, 98.13%, 91.79%, 89.21%, and 90.48%, respectively. Comparisons among the various model architectures are provided by the study. Researchers and practitioners can use the findings to pick suitable transfer learning models for tasks related to medical image categorization.