A 96-active-qubit benchmark on IBM’s Marrakesh processor produced 30 of 30 regional matches to intended reference patterns across five hardware runs, an IEEE-presented study said.
More than 6,000 gate operations—including 1,241 two-qubit CZ gates—were tested without error correction or post-selection, and the key HOG metric averaged about 65% versus a 50% no-preference baseline.
Three control circuits also generated structured outputs, but each matched its own ideal reference best, suggesting the method can distinguish the intended circuit result from other nonrandom patterns.
A separate 16-qubit experiment converted the dominant measured state into a “PING” command that authorized a standard UDP packet between two classical computers, creating an auditable link from quantum output to classical action.
The work does not demonstrate a full distributed quantum computer or network, but argues recognizable, validated quantum outputs should be a basic readiness test for future hybrid quantum-classical workflows.