IMO both approaches have benefits.

Sampling can more closely represent the instrument as it was recorded, at a specific point of time, a realistic snapshot as it were.

Modelling can smore broadly represent the instrument as it is being played, (nuances, resonances etc), essentially more about the dynamic representations of the sound over time.

Personally I prefer sampled methods, but neither are close enough to piano to replace it, so all in all both are substitues for the real thing.

But neither of these though would be my priority top in regards to learning, more like:

1. As many keys as you can get, up to 88
2. At least semi-weighted or preferably fully weighted
3. Foot pedals, damper, sustain etc
4. Sample or Modelling
5. Proper stool, Headphones

There's a thread here that might help, with babu bearing source of some of the references

IMO both approaches have benefits.

Sampling can more closely represent the instrument as it was recorded, at a specific point of time, a realistic snapshot as it were.

Modelling can smore broadly represent the instrument as it is being played, (nuances, resonances etc), essentially more about the dynamic representations of the sound over time.

Personally I prefer sampled methods, but neither are close enough to piano to replace it, so all in all both are substitues for the real thing.

But neither of these though would be my top priority in regards to learning, more like:

1. As many keys as you can get, up to 88
2. At least semi-weighted or preferably fully weighted
3. Foot pedals, damper, sustain etc
4. Sample or Modelling
5. Proper stool, Headphones

There's a thread here that might help, with babu bearing source of some of the references