Originally posted by Marc Sabatella There's no getting around the fact that most people who have compared subjectively seem to prefer the 3.5 (as evidenced on Stan's site), but Yoshihiko's numbers do kind of suggest otherwise. My own quick haphazard comparison using my 2.8 and a borrowed 3.5) suggests that whatever difference there is, it's too small for me to care about.
The K (as opposed to M) version does better in the numbers, and indeed, most people call that the best of the bunch. But it's harder to find, usually more expensive - and also quite a bit heavier. Personally, if I were going that route, I'd hold out for the even better and slightly faster (yet somehow lighter) K30/2.8.
But as cheap and easy to find as the M's are - both f/2.8 and f/3.5 versions - and as good as they both are already - I'd be more inclined to just get one of those. If you're *really* looking to step us, there are always the M and A 28/2's.
I noticed the numbers as well. if we are going to go by the numbers, the M version would be the better lens. however, the question is, would those numbers show a much better image rendering or image illustration?
as was discussed before, the K28/2 is supposedly a wide portrait lens where sharpness is not it's purpose that is fast and has a shallow DOF. it would be the opposite of the K28/3.5 which is a wide sharp lens where it is primarily good for subjects of a more solid DOF. I would say the Yoshihiko numbers are pretty consistent to how the lenses are performing in real world.
this might be the reason why the K28/2 is much preferred over the M28/2. the M28/2 is as fast and maybe sharper than the K28/2 all over. but this made the lens less desirable because it doesn't have that certain image effect that the K28/2 has. the floating element could be the main factor here.
if I ever get some mad money, I might buy the CZ version instead since the K28/2 is really pretty scarce and I had let one slip by recently at a pretty decent and affordable price.