Model compression by using Sequential API twice?

I have a sequential model called model and when I do tf.keras.sequential(model), there was a huge drop in the model size without any change in its accuracy . Can someone explain what’s happening here?

Please provide standalone code to reproduce the issue? Thank you