I would like to include a theorem from a paper, say Alice et al. [12], in my own manuscript.
Which option is the most readable one?
Theorem 6 [[12]]. Every even number can be divided by two without remainder.
Theorem 6 [Alice,Bob and Charles [12]]. Every even number can be divided by two without remainder.
Theorem 6. Every even number can be divided by two without remainder [12].
Will you ever cite any source for a reason other than to attribute a theorem to it? There are several popular conventions for how to do this, and you may as well use one consistently, including for the citation of theorems. The second option you've listed is especially strange because people tend to write [12] or Alice 1984a, depending on whether they've bothered numbering the sources globally; one doesn't number them and then name them anyway, outside of the list of the references.
So the real question is whether to do things the first or third way you suggest. Either way, I advise a slight tweak wherein you make clear which of the source's theorems you're borrowing. For example, option 3 then becomes
Theorem 6. Every even number can be divided by two without remainder (viz. Theorem 34b in [12]).
