max C(n,2) = 28
max C(n,3) = 56
more variables than current IBM chips hold (156): needs decomposition into chip-sized subproblems or a sparser reformulation, see Part 2 of the tutorial
clean shots per iteration
λ = 0.0840, 92% of runs clean, needs ~≥256 for this size
| terms Nint = pairs + triples | — |
| native 2q gates N2q = pairs·2 + triples·4 | — |
| routed gates (× r) | — |
| λ = routed · p2q | — |
| clean fraction e−λ | — |
| clean shots | — |
λ counts two-qubit gate errors only; readout and decoherence add on top, so treat this as a floor.