IBM FEZ · 23 JUL 2026

150 qubits.
599 QPU seconds.
Zero advantage claims.

Explore a hardware-native Ising study across QAOA depths, mitigation choices, and an independent p5/p6 confirmation—anchored to an exact classical solution.

Open dataset ↗ Reproduce on GitHub ↗
150active physical qubits
15 jobs547 PUBs2.09M shots
Exact classical865 / 873HiGHS · MIP gap 0
Replicated best mean0.742274p=6 · 22 replicates
Replicated best sample0.915607p=6 · 90,112 shots
QPU allowance used599 / 600 ssuccessful jobs only

INTERACTIVE EVIDENCE

What changed as the circuit got deeper?

Mean and best-sample approximation ratio

Mean quality improves through p=6, then plateaus. The best individual sample in the initial sweep occurs at p=5.

THE RESULT

Depth helped—until it didn’t.

The p=6 mean survived replication and improved from 0.739050 to 0.742274. Extra depth beyond p=6 did not improve the initial mean.

+10.3%relative mean gain from p=1 to replicated p=6

TRUTH BOUNDARY

  • ✓ Hardware execution verified
  • ✓ Exact classical baseline verified
  • × Quantum acceleration not measured
  • × Quantum advantage not claimed

For this instance, classical HiGHS remains the production winner.