I am not going to recommend a make, though I have Focal Utopia’s and what I use does sound excellent with them, but might push your budget because you don’t tend to find them used, for good reason.
What I do suggest is that whilst good design of experiment says "change one item at a time" the issue in most hifi cases is you will not see the full improvement until you have removed, or at least balanced out the bottlenecks though the system. I had bought an upgraded AES/EBU (within a dCS Vivaldi system) and a pair of same make XLR’s for variable output DAC to Vitus Power AMP. But was waiting on my Speaker cable upgrade and in the interim had been lent a pair of speaker cables from the same manufacturer but 2 grades down from the AES / XLRs.
Whilst the speaker cable made some improvement of the previous I was disappointed (less experience at that point) in the difference of swapping the AES/XLR’s in and out with against the ’basic’ Van Damme cables that had come with the dCS, However, when the speaker cable (and also actually a clock cable) were upgraded to the same level I could then hear a significant improvement.
So whether you home trial or find a dealer who has Sopra 2/3’s for an initial trial I would make sure you are getting them to either step up through a full system cable up grade, in either case start at base with your cables or same level if at dealer and then:
1 Increment speaker cable
2 Increment XLRs
Repeat loop until no further improvements, or you don’t trust yourself with the price increase on the next one.... but you will still have to know.
Of course according to naysayers it will never change and according to the (sometimes correct) diminishing returns fans it will reach an asymptote.
However, one situation that can confound the asymptote is when there is a change in for example the dielectric in an upper level cable. Then that can (and IMO and all who have experienced it in my system does) produce a bigger change than the still appreciable increment of those lower down in the range.
Happy testing.