The cut-off for inclusion in the model hasn’t changed in the past few years: 100 observations, at least 50 research grade. See https://www.inaturalist.org/pages/help#cv-taxa
If enough photos are present at the genus level, but not enough photos for any of the descendent species, then the genus will be placed in the training set. We call this approach a “leaf model.” 
More details are available in Ken-ichi’s TDWG keynote from last year:
https://www.youtube.com/watch?v=xfbabznYFV0