I remember discussions on that in other forums... At Musikerforum, some synth specialists thought Roland SuperNatural Acoustic sounds were indeed modeling, while Yamaha‘s SA2 voices were just based on „scripting“. Scripting is largely using various chunks of samples in flexible combinations depending on how tje sound is played on the keys. Modeling changes the actual waveform in real time. That‘s quite a difference.
If you think of how you can gradually (!) change the sound of a piano in the Integra with the sound character control (not like just changing how mellow or bright it is, but much more substantially), this is an indicator that there is real modeling involved.