The SCScore model assigns a synthetic complexity score between 1 and 5 to a molecule. The score is based on the premise that published reactions, overall, should exhibit an increase in synthetic complexity. The model has been trained on 12M reactions from Reaxys.
The standalone numpy model is defined in scscore/standalone_model_numpy.py