Just to add, this is an RF amp and the FET's run at 137kHz, 50V at about 20 to 25 Amps at the moment, on FET's that are a bit near their limit.
Okay, well, if you are driving the MOSFETs at 137kHz to get a 137kHz output then the previous advice applies [edit - delete irrelevant commentary on class of operation since I'm not entirely sure whether 137kHz is modulated or not].
So, for example, the MOSFET you referenced earlier has a Qg of 100nC and the usual rule of thumb for smps design is to make the transition time (ton or toff) 1% or less of a switching period, or 73ns in this case. 100nc/73ns = 1.4A, which is within the capability of the IR2110 you also referenced earlier.
But if this is a true Class-D amplifier, in which a lower frequency signal (ie - 137kHz) PWM's a higher frequency carrier followed by low-pass filtering to recover the modulating signal, then your switching frequency might need to be 600kHz or more to allow for a simple LC low pass filter on the output. That means a transition time in the sub-10ns range, which will require a peak current that will likely exceed the ampacity of the gate bond wires and/or die metallization, while the stray inductance of the package alone will conspire to prevent slewing that much current that quickly with just 10V of drive.
It is possible to drive the gate resonantly, or use a matching network to transform the impedance, but this is pretty advanced stuff.