Google Demonstrates Quantum Contextuality on Willow Chips, Reporting Results Up to 105 Qubits

GOOGL

·

Google says it has demonstrated at much larger scale one of quantum mechanics’ most debated features — contextuality — on its Willow superconducting quantum processors, reporting task results beyond specific classical limits on problems involving as many as 105 qubits. In a company blog post published Wednesday, Google framed the work as both a test of quantum foundations and a potentially practical way to measure hardware quality.

The clearest result for nonexperts comes from the Mermin-Peres “magic square” game, a standard coordination task used in studies of contextuality. According to Google’s Nov. 19 blog post and the underlying arXiv preprint, the company’s processors achieved an average winning probability of 0.9830, or about 98.3%, compared with the best classical average of 8/9, or about 88.9%. The paper says all nine distinct variants of the game were each played 100,000 times. The experiments used Willow-series chips including a 72-qubit processor, with results reported on a 71-qubit sub-grid, and a 105-qubit processor.

Contextuality is the idea that, in quantum mechanics, the outcome of a measurement can depend on which other compatible measurements are performed alongside it. That differs from simple classical models, where measurement outcomes are treated as fixed in advance. Physicists and computer scientists care because contextuality has long been argued to be one of the resources behind some forms of quantum computational power, including in influential theoretical work by Robert Raussendorf in 2013 and by Mark Howard and colleagues in Nature in 2014.

The arXiv preprint, titled “Quantum-Classical Separation in Bounded-Resource Tasks Arising from Measurement Contextuality,” was posted Dec. 1, 2025, and lists Pedram Roushan and Michel Devoret among its corresponding authors. Its central claim is that “using a superconducting qubit processor, we show that quantum contextuality enables certain tasks to be performed with success probabilities beyond classical limits.” Beyond the magic-square game, the team also tested an N-player GHZ game and a 2D hidden linear function problem, both commonly studied in quantum information. The paper additionally reports genuine GHZ-state entanglement up to 45 qubits without readout-error mitigation.

Google and the paper’s authors present that as more than a foundations milestone. They argue that contextuality-based tasks could serve as a sensitive benchmark for coherent errors — the systematic control errors that can build up in quantum hardware — as well as overall processor performance. In its blog, Google said its qubits “won a well-known coordination game 98.3% of the time, beating the 88.9% classical limit, and outperformed constrained classical strategies in tasks spanning up to 105 qubits.” That framing matters because quantum computing researchers are still looking for tests that reveal not just whether a chip works, but how errors behave across many qubits and many operations.

The result is notable partly because earlier experiments probing contextuality were generally much smaller in scale, while Google is claiming a processor-level demonstration across dozens to more than 100 qubits. But the framing needs care. The work was available first as a December 2025 arXiv preprint rather than a newly released journal paper, and its classical comparisons are not blanket claims that all classical methods fail. Instead, the reported separations are against bounded-resource or otherwise constrained classical strategies on specific tasks. That makes this a more precise story about scaling a foundational quantum effect — and about turning it into a hardware benchmark — than a broad claim of general-purpose quantum advantage.

Tags: #quantum, #computing, #contextuality, #google

Stocks: GOOGL