Overview
- On Thursday, July 30, 2026, IBM and the University of Chicago reported a logical‑qubit run that encoded 70 logical qubits, executed 2,415 logical two‑qubit gates and 468 logical T gates, produced logical error rates about ten times lower than physical rates, and completed a hard sampling task in roughly 15 minutes.
- Also on July 30, IBM and Qedma said they used Qedma’s QESEM error‑mitigation software to observe long‑lived dynamics on up to 74 qubits, and they compared those results with classical simulations from RIKEN and BlueQubit and hardware checks on Quantinuum that showed classical methods disagreed at these problem sizes.
- IBM and Algorithmiq reaffirmed a heterogeneous‑matter benchmark first posted eight months earlier and reported that no classical method has reliably reproduced results across the full regime, while releasing new benchmarking code to let researchers test claims openly.
- The teams combined three trust-building approaches—large‑scale logical encoding, software error suppression and deliberate noise‑manipulation—plus cross‑hardware checks and public release of circuits and data to provide measurable validation when direct classical verification is infeasible.
- Researchers say the work makes quantum processors more usable as scientific instruments for hard physics and materials questions and that ongoing community benchmarking and independent checks will be needed; observers also note the same advances could inform future cryptographic risks and post‑quantum planning.