The first one allows to send a value for the note-off velocity.

The second one allows to use running status, i.e., saving one byte (90 2B 72 2B 00).

The 8x method is easier to read by humans.  It also allows note-off velocity, but you need a synth that supports the note-off velocity to make use of it and hear any results.  It affects what happens to a note when it is "released".

The 9x method is easier for old MIDI hardware to use, since the velocity is always 0, and it allows the use of running status to reduce the number of bytes that need to be transmitted through a MIDI interface. (the 9x byte can be excluded if the previous event was also of the same 9x type.  So a note on 9x 05 75 immediately followed by a note off 9x 05 00 would only need to transmit 9x 05 75 (00) 05 00 [with (00) the delta time between the two]).  So (2) 3-byte commands are reduced to a 3-byte and a 2-byte command.  And if you have an entire line consisting of just note on and note off, the byte reduction adds up quickly, as the first note on would be 3 bytes, and all following note off and on would be 2 bytes each until the next command of a different type.