Different types of thoracic diseases or lung diseases have become life-threatening for common people. An early diagnosis of lung illness is crucial. The extremely contagious Corona Virus Disease, also known as COVID-19, has a significant negative influence on world health. Because the virus spreads through contact with humans, it has had a disastrous impact on our lives and put tremendous demand on the global economy and public health system. The rapid spread of COVID-19 can be checked if detected early and provided medical treatment accordingly. According to WHO the leading causes of death worldwide are lung illnesses and pulmonary abnormalities. Many researchers applied Deep learning techniques to detect thorax disease; most of them revealed VGG (visual geometry group-based neural network) as the best model. In this study we proposed CNN model that has classified Covid-19, lung opacity, pneumonia, and normal from chest X-ray images. The proposed model has shown better results as compared with VGG19, ResNet50, MobileNetand InceptionV3. In this work we have used accuracy, recall values, precision value, and F1 score to compare the proposed model with other state-of-the-art models. We have received overall accuracy as 96%, which is the best as compared with other models tested. The proposed model’s correctness has also been contrasted with that of a more current model.