Scikit-learn models might take a while to train. That's why it's preferable to store them on disk such that you may be able to re-use them elsewhere.

The standard method of doing this in scikit-learn is to use joblib to store a pickle file.