The COVID-19 pandemic stands as one of the most formidable diseases ever recorded, emerging suddenly and wreaking havoc not only on social and economic structures but also on global public health. A significant challenge has been posed in accurately detecting COVID-19 and distinguishing it from other respiratory illnesses, especially through X-ray images that contain subtleties not easily noticeable to the human eye. Deep learning models have emerged as robust tools to address this challenge. These algorithms provide a more detailed analysis of images than conventional techniques, significantly enhancing the accuracy of disease detection based on imaging. This study focuses on employing transfer learning methods, specifically utilizing VGG16, VGG19, DenseNet201, EfficientNetV2B3, MobileNetV3Large, Xception, and Inception models. These models were systematically evaluated and compared, with DenseNet201 emerging as the most suitable model, achieving a remarkable accuracy of 97%. This surpasses other state-of-the-art models, offering a more efficient approach to early COVID-19 detection.