As a matter of fact, we published just this year on a novel algorithm using just two qubits! The trick is that this algorithm is in some sense "generic" over arbitrary data structures, so we were able to claim victory by tackling the simplest possible case. It was still a challenge in practice, since it required a lot of sneaky tricks and record (to our knowledge) two-qubit gate fidelities.

We're also working with larger devices, but I'm not sure if I'm supposed to comment on those right now :)