This is an unusual situation in mathematics: I'm not sure if I've ever seen a singly claimed theorem in a multiply authored mathematics paper except when the theorem has its provenance in explicitly mentioned earlier work of the single author. (I would be interested to see an example.) I'm pretty sure there is no "standard" answer.
One idea would be to bail out of listing either author's name: you could just say "Our argument is based on [7, Theorem 3.1]...." This is not ideal: I think that when you cite someone's work in a critical way then their name should appear in the text itself rather than be pointed to / abbreviated in the bibliographic citation. But this is not a hard and fast rule, so far as I know.
I suppose that if the paper itself says the theorem is due to Smith alone and not Jones-Smith, then you should attribute it that way in your writing. Thus all of your suggested phrasings seem appropriate to me. Readers who see "theorem of Smith [7, Theorem 3.1]" and then flip to the end to find a paper of Jones-Smith may be a bit surprised...but then they'll read the paper and see that you've reported the attribution as Jones and Smith themselves did.
If this is a really famous theorem then the community at large -- or even different portions of the community -- may have its own feelings about how to refer to it. (A vaguely similar instance in contemporary mathematics is that some people speak of Maynard's Theorem and others speak of Maynard-Tao...) In this case, by saying one thing rather than another you may be signalling some kind of political allegiance / personal fealty....Such issues are beyond the scope of this answer.