One of IBM's "quantum advantage" experiments is reproduced on ordinary computers
IBM and partners posted three papers in July claiming quantum advantage. A new preprint reproduces one of them, a 70-qubit experiment, on conventional graphics processors in about 37 minutes.
In July 2026, IBM and its partners (Algorithmiq, Qedma and the University of Chicago) posted three papers describing experiments on IBM's Heron processors that they put forward as quantum advantage: tasks a quantum computer does better than any classical one.
On 13 August, a separate group posted a preprint that reproduced one of them, the 70-qubit experiment with the University of Chicago, using 256 high-end graphics processors in about 37 minutes. None of the three original papers has been peer-reviewed yet.
What it means: this is how advantage claims get tested. A classical reproduction doesn't make the experiment worthless, but it does mean that particular task isn't evidence of advantage. The other two claims still stand as candidates. IBM has said it expects community-confirmed advantage by the end of 2026, and our IBM profile tracks that promise as due.
Does this change the rankings? No. Rankings only change when QubitBenchmark publishes a checked result in its monthly data release.
Sources
- Classical simulation and design frontiers for IBM's doped Clifford sampling experiment, arXiv preprint 2608.13110, 13 August 2026.
- IBM's three quantum advantage claims, fact-checked, PostQuantum.com, 1 August 2026