Hardware
Google's Willow chip beats classical limits in contextuality tests, Science paper reports
Google Quantum AI researchers reported in Science on Nov. 19 that the Willow processor beat classical limits in tests based on quantum contextuality. In a magic square game, Willow won 98.30% of rounds across nine versions played 100,000 times each, against a classical maximum of 88.89%. A Bell-Kochen-Specker inequality that non-contextual models cannot push above 4 measured 5.590 on Willow, against a quantum maximum of 6. A GHZ parity game stayed above the classical win rate with up to 71 physical qubits, with entanglement verified across 45 qubits, and a hidden linear function problem was run on grids of up to 105 physical qubits, with success falling from about 1 to about 0.6 at full size.
Why it mattersGoogle says the quantum ceiling on these tests is fixed by physics, so they could serve as a benchmark that does not shift as hardware improves. The paper's caveats are that noise grows with qubit count and narrows the quantum lead, and that the classical bound for the 105-qubit problem is conjectured, not proven; a preprint posted to arXiv in December 2025 is not mentioned as peer reviewed in the article, though the Science paper is.