It uses the 2S or “6V” led. I call it 2S as it means “2 in series”, this is because XHP50.x (along with XHP70.x) emitters feature 4-dies usually arranged in 2 groups connected in series, each one consisting in 2 dies connected in parallel. With special boards the dies can be arranged 4S (4S1P) or “12V”, these boards are available from SinkPAD at some places like Mountain Electronics.
If you want to try an Osram buy an SST40 version. SST40 versions employ a ∅22mm, 6A maximum linear driver for 1S or “3V” leds: 4-modes or 12-groups. With up to 6A it is a very adequate driver for the CSLPM1.TG or CULPM1.TG, “2mm² white flat”, without going overboard with driving current, which in my lightful opinion is what happened with the new “8A single cell buck driver for CULPM1.TG”: it's somewhat overcooked.
Additionally, and concerning (XHP50.x/XHP70.x) 2S or 6V emitters, I can see the new drivers for XHP50/70 emitters (XHP50.x 12-groups, XHP50.x 4-modes, XHP70.x 4-modes and XHP70.x 12-groups) with pretty much “stock” driving currents. Considering the availability of more powerful drivers for 1S or 3V leds, I find this a bit underwhelming. Of course, this is just an opinion of someone always looking for an extra oomph. So, I recommend installing the XHP70.x boost drivers with/for XHP50.x emitters, i.e. if you buy an XHP50.2 S21B, you could give Simon a tip and ask him to put an XHP70.x driver in it (if he accepts, mind you). This of course is a call for more powerful boost drivers; XHP50.x and XHP70.x emitter tests prove they can handle twice the stock maximum driving current guaranteed. Some test examples: XHP70.2 P2 4000k Output test by Texas_Ace, early XHP70 test by djozz (up to 11.5A), XHP50 Output test by Texas_Ace and earlier XHP50 test by djozz (both maxed out at ≈9.25A).