There is a lot of "black magic" that goes into creating models that do not have discontinuities in their I/V characteristics. I have seen Spice get stuck with a third-party model of a 1n4007 where just replacing that with the standard Diode model (the one built-into the STD lib) will make the simulation run infinitely faster.
I have seen this many times before where LTspice struggles with MOSFET models. If the simulation hangs as soon as the MOSFET tries to turn on, then this is normally indicative that the FET model is either too complex (and LTspice is struggling with it) or the model is wrong.
LTspice uses a simplified model for all of its MOSFETs. It models ON resistance and gate charge accurately, but pretty much ignores many of the other parameters in an effort to get the simulation to work fast.
As Alec suggests, I strongly recommend you use the LTspice models. Pick one with a similar VDS, QG and RDSON to the one you intend to use and run the simulation. It should run a lot faster. I would avoid trying to edit the model yourself - this is plagued with headaches.
Same applies to many other components in LTspice (diodes, bipolars), but MOSFETs are the worst offenders for causing your simulation to hang