If you can view the data at the service provider, Nathan has the best answer. I believe we're saying the same thing, but to restate: full siblings will be half-identical (HIR: half-identical region) on 50% of their DNA and also fully-identical (FIR) on an additional 25%. If we're looking at the "standard" total centiMorgan evaluation that most services use (GEDmatch, FTDNA, Ancestry, MyHeritage), which assumes (not quite accurately, but then cM calculation itself is only approximately accurate) the whole genome is 6,800cM, so that would equate to 1,700 HIR plus 850 FIR. As Nathan notes, half-siblings--assuming no immediate-generation pedigree collapse--should show no meaningful fully-identical regions at all. The only other relationship that will show significant FIR (well, other than identical twins, of course) would be double first cousins, and even then it would only amount to about 106cM, or around 1.6%.
GEDmatch provides this information, and you can at least view it as a graphical display (but you have to eyeball the values) at 23andMe in the Family Traits chromosome browser.
Back to the aggregate cM amounts, it's worth noting that 23andMe does things a little bit differently than everyone else. They compute the 22 pairs of autosomes as equalling 7,074.6cM instead of 6800cM. They also are the only company that adds the X chromosome into the total calculation. After including the single X, the male total cM at 23andMe is 7,256.8; with the two X chromosomes, the female total is 7,438.6cM. So unless you can view 23andMe results elsewhere, like GEDmatch, you can't get an apples-to-apples comparison with a test-taker from another company, and the 23andMe reported sharing will almost always be higher, at least for recent relationships.
Using the 6,800cM model, the theoretical average sharing for full-siblings would be 2,550cM; for half-siblings, 1,700cM.