Okay, I have to speak up here, since modelling is my trade. 🙂
First of all, there is no modelling without sampling, because sampling delivers the data from which the models are built.
Secondly, the most evolved and refined version of sampling would bring a result not distinguishable from recorded sound from a piano. It can go no better. But since such a result would make the digital performance equal to the acoustic performance (recorded, not experienced live), that would also be attaining perfection. Modelling, if done right, would be able to reach that same level of perfection. It would also be able to "go further". But since this "further" would be parting ways with the recording of the acoustic piano, it would also be less perfect.
Now, it could be argued that there is no reason why it should not be possible, through artificial means, to create something better than the sound of the acoustic, and I, for one, would have to agree. But history tells us of many such attempts (theremins, and synthesizers, anyone?), and to date none of these have been able to knock the acoustics off their throne. The artificial ones come and go, and live only forever in the memory of resigned nostalgics.
Will the artificial sounds win out in the long run? I don't know, but I'm not holding my breath, waiting for it. 🙂