@Tim 'mithro' Ansell there's two separate qualities. One being the accuracy of the model when compared against measured data in very controlled conditions. Secondly, the numerical performance with a simulator when trying to design circuits efficiently. It's often possible to have one without the other although of course we strive for both. I would guess that it's not possible to move from a binned model to a continuous model without affecting the numerical performance of the model, I'm hoping the change is positive. But I'm not a spice expert so I could be wrong.