pb range 0-16383.  In 14-bit-land thats:

00000000000000 (binary, 14 zeros)

11111111111111 (binary, 14 ones)

Just orienting myself here - not stating anything earth shattering.

The previous Montage classic would have had a pitch difference between 5500 (decimal) which is 01010101111100 and 5550 (decimal) which is 01010110101110 because with Montage Classic the bolded 7-bits of least significant bits are ignored and you can see the unbolded high side has a different value (0101010 vs 0101011).  The Montage Classic would not have a different pitch between 5550 (decimal) which is 01010110101110 and 5600 (decimal) which is 01010111100000 because the unbolded numbers are the same and, again, the LSB bolded values are ignored in Montage classic.

As an aside - this isn't a MIDI 2.0 thing.  MIDI 2.0 gains over double the amount of resolution (32 bits).   Montage Classic was "throwing away" MIDI 1.0 resolution.

Jason's reply shows up for me for some reason. But replies in other threads are still broken.

I didn't say it previously but since you have reported audible pitch differences between 5550 and 5600, it appears the Montage M is using more bits of MIDI 1.0 than Montage Classic does.

I don't know if the MIDI2.0 workbench allows for you to inject MIDI 2.0 but that may be a tool to send MIDI 2.0.  Maybe using a mac.  I know Microsoft is using the tool and their drivers to do things although I guess there hasn't been a (driver) release that was hinted 7 months ago @ NAMM.  There was a Spring 2023 release but maybe it's too limited for USB device use.

If pitchbend stepping is not showing on the M, that is a good sign.

For me the real test is to A to B the same FM-X sound on my DX7II and Montage.  I can almost instantly tell (by ear) if there is stepping.  I would trust my ears more than the MIDI data, mainly because I don't dig that deep into MIDI.  I just know when I hear the stepping, even if subtle, it is not pleasing, when my 35 year old synth is silky smooth with absolutely consistent zero audible stepping...

For me the real test is to A to B the same FM-X sound on my DX7II and Montage.  I can almost instantly tell (by ear) if there is stepping.  I would trust my ears more than the MIDI data, mainly because I don't dig that deep into MIDI.

You are right. It isn't just about what data is available or what precision it has but. It is more about what the instrument actually does with that data.

Further tests with MIDI-OX show that, on the M8X, moving the PB wheel ever so slowly produces the entire range of 128 values for the MSB of the precision. 

Since a -48 setting only uses fewer than 128 they are all covered. On the M you can produce all 48 and my ears tell me there are some in between also. So nothing is 'stepped' as far as I can tell - YMMV.

Using MIDI thru doesn't really indicate what the instrument is doing internally. You can't get/put MIDI 2.0 data through the previously existing ports - they added a new port to the M's specifically for MIDI 2.0 support. But the MIDI 2 workbench isn't available yet for Windows and even the version for Linux is in Alpha test at best.

There are some noticeable precision changes on the M. The knobs are 0-1023 now and not 0-127. The assign curve ratios are from -128 to +127 and not -64 to +63.

If pitchbend stepping is not showing on the M, that is a good sign.

For me, the real test is to A to B the same basic DX/FM-X sound on my DX7II and Montage.  I can almost instantly tell (by ear) if there is stepping or not on the Montage.  I would trust my ears more than the MIDI data, mainly because I don't dig that deep into MIDI, and do most things like this by ear.  I just know when I hear the stepping, even if very subtle, it is not pleasing to my ears, especially when, in almost immediate direct contrast, my 35 year old synth beside the Montage is so silky smooth with absolutely consistent zero audible stepping...