Quantum advantage through contextuality: Google's experiment on the Willow processor

Edited by: Svitlana Velhush

Tweet not found

The embedded tweet could not be found…

On the Willow quantum processor, the Google Quantum AI team achieved success in coordination games where classical strategies are limited to 88,9 percent, while quantum ones reach 98,3 percent. The results, obtained without any communication between the participants, violate classical limits and demonstrate a quantum advantage based on the contextuality of measurements.

The experiment was conducted by Google Quantum AI researchers in the USA on the superconducting Willow processor with a qubit count of up to 105. They implemented several tasks: a two-player "magic square" game, an N-player Greenberg-Horne-Zeilinger game with up to 71 qubits, and a hidden linear function problem on a two-dimensional lattice with up to 105 qubits. The paper was published in the journal Science on 19 November 2026 and underwent peer review.

Contextuality means that the outcome of measuring one property depends on which compatible measurements are performed simultaneously. It is as if the answer to the question about the color of an apple changed depending on whether you simultaneously ask about its shape. In classical physics, properties can be assigned to objects in advance, but in the quantum world such pre-assignment is impossible without contradictions.

In the "magic square" game, two players must fill a table so that the sums of the rows and columns match the given values. Classically, the maximum win is limited to 88,9 percent, but quantum measurements on entangled qubit pairs made it possible to achieve 98,3 percent. An additional test using the Bell-CHSH inequality gave a value of χ = 5,590 against the classical limit of 4.

These tasks serve as a new benchmark for evaluating quantum hardware: they reveal subtle errors that random circuit tests miss. The advantage persists when scaling up to 105 qubits, which opens the way to verifying contextuality in real algorithms.

The results were published in Science on 19 November 2026 and confirm that contextuality works in multi-qubit systems. They give physicists and engineers a precise tool for comparing processors without relying on classical supercomputers.

Quantum processors can now be tested according to the laws of quantum mechanics itself, rather than by the speed of overtaking classical machines.

1 Views

Sources

  • Quantum-classical separation in bounded-resource tasks arising from measurement contextuality

  • research.google

  • blog.google

Comments

Did you find an error or inaccuracy?We will consider your comments as soon as possible.