On the eve (almost) of Canon launching its 32mm f/1.4 lens for EF-M mount (i.e. 1.6 crop factor APS-C), let me point out that in terms of field of view and achievable shallow depth of field, this is roughly the same as a 50mm f/2 on full frame.
I thought it would be useful to write a brief reference for folks to refer to when comparing lenses with different imaging circles (e.g. full frame vs. APS-C), giving specific values for popular focal lengths and apertures.
So in this vein, a 35/0.95 APS-C lens (e.g. Mitakon) is roughly like a 50/1.4 on full frame, and a 50/1.1 for APS-C (e.g. Kamlan) is similar to 85/2 on full frame. So the sometimes affordable 85/1.8 lenses for full frame cameras don’t get you into much better territory than 50/1.2 on APS-C, although you may by this route obtain autofocus at a similar price but at the additional cost of needing a full frame body.
Also note that light equivalence is already taken into account – there’s almost a stop difference between f/1.8 and f/1.2, with a little bit of wiggle room for manufacturer’s specs (aka cheating).
Once upon a time, it used to be true that digital sensors could not work at shallow depth of field very well as the angle at which light would typically fall onto the sensor was typically shallow as well and many of these shallow rays would be deflected off the surface of the sensor, avoiding detection. So on the assumption that that’s still true of modern sensors, working with an f/2 lens on full frame would be preferable to working with f/1.4 on APS-C. One might assume that sensor design has markedly improved in this aspect since mirrorless cameras have to cope with even shallower angles – however, note that mirrorless lenses are typically a little narrower at the aperture and recent lens designs have been very long and heavy, giving an overall less acute angle of incidence, which might be a hint as to the remaining challenges in this area.
If sharpness is what we care about, a full frame camera is generally preferable as the format allows for greater manufacturing tolerances in lenses, and the same lens will, generally speaking, give a sharper image on full frame than APS-C, depending on the sharpness fall-off towards the corners.
And one more thing – if we compare the front element sizes of, say, a 32/1.4 (22.9mm) and 50/2 (25mm), we don’t see a lot of difference, so we might suspect that in terms of the torque needed for autofocus to move elements inside such lenses, there may not be a lot to be lost or gained. So there may not be any particularly convincing excuse for a 35/0.95 lens to not have autofocus, unless we take into account the generally smaller build and smaller battery of a typical APS-C camera compared to a full frame model.
But getting back to the equivalence issue, it’s fair to conclude that most “fast” lenses for APS-C sensors offer nothing particularly revolutionary, this being even more true if they have to be focused manually.