Superconducting processor shows contextuality in tasks up to 105 qubits
Google Quantum AI reports measurement-contextuality tests on its Willow processor, including a shallow hidden-linear-function circuit on grids of up to 105 qubits.
Contextuality is the idea that a quantum measurement's outcome can depend on which other compatible measurements are made alongside it, rather than on values fixed in advance. It has been proposed as a resource for some quantum algorithms, but until recently it had mostly been shown in small systems of a few particles.
In a Science paper published 19 November 2026, Google Quantum AI reports tests of that property on its superconducting Willow processors. In an accompanying research blog, the authors say their qubits won the magic-square game in 98.30% of rounds (0.9830 ± 0.0001), after playing all nine versions of the game 100,000 times each. The best classical strategy, they note, can win at most 88.89% of the time. They also report a state-independent Kochen–Specker–Bell score of χ = 5.590 ± 0.008, compared with a classical bound of 4 and a quantum limit of 6.
To ask whether the effect still shows up at larger scale, the team ran a shallow hidden-linear-function circuit on grids of up to 105 qubits. The same blog says the probability of success fell from about 1 to about 0.6 as the system grew to the full processor size, remaining above a conjectured classical bound. The experiments are a hardware demonstration and a proposed benchmark, not a commercial application, and they are not a claim that a useful end-to-end algorithm has beaten the best classical supercomputer.
Why it matters
If contextuality is one of the ingredients that gives certain quantum circuits an edge, then being able to measure it on a processor with more than a hundred qubits gives engineers a concrete yardstick. The authors also say the tasks exposed subtle coherent errors that random-circuit tests can miss.
Looking ahead (speculative)
As processors get quieter, scores on the same tests should move closer to the physical ceilings (a perfect win rate, or χ = 6). That would make the benchmark more discriminating. Whether those gains will translate into practical algorithms is a separate, still open question.
Sources
Sources
- Kumar et al. (Google Quantum AI and Collaborators), Quantum-classical separation in bounded-resource tasks arising from measurement contextuality, Science, 19 November 2026
- A game you cannot win unless you have a quantum computer, Google Research, 19 November 2026
- Testing quantum contextuality at scale on Google processors, Google, 19 November 2026