No such animal.
Any programs that purport to do this can't accurately tell a human voice of a certain frequency from a saxophone playing the same frequency. If the program can remove a 3000 Hertz human voice note, it will also remove a 3000 Hertz saxophone note.
This is further complicated if the vocal and the instruments are in the same channel. If they are in different channels you can have some limited success with programs such as Audacity.
Keep your expectations very low.
If the vocal is entirely in its own channel, then you can have success by deleting everything in that channel---but that is RARELY the case. In most recordings, the instruments and the vocal are NOT in completely separate channels, which causes the problem you are encountering.