Being unchanged for decades means that the training data should provide great results even for the smaller models.