Google's Willow chip wins quantum coordination game 98.3% of the time on four qubits

Editorial illustration: Four glowing spheres connected by crossing light trails on a metallic platform beneath a three-by-three grid of alternating blue and cream tiles.

In brief

  • Willow won a four-qubit coordination game 98.3% of the time, against an 88.9% classical maximum.
  • The result drew on 900,000 trials; hardware noise kept it short of the quantum ideal.
  • A 105-qubit task reached roughly 60% success, failing about four in 10 runs.
  • Contextuality tests catch subtle hardware errors that randomized tests can hide, Google says.
  • The study isn't commercial quantum computing, and it doesn't beat every classical machine.

How the magic square game works

The first test used the Mermin-Peres magic square game, in which two players fill intersecting parts of a three-by-three grid without communicating. Their answers have to obey conflicting row-and-column rules and match at the shared square. A prearranged classical strategy can't succeed on more than eight of nine assignments (that's the 88.9% ceiling).

Entangled qubits can theoretically win every round, according to the researchers. Willow won 98.3% across 900,000 trials, and unavoidable hardware noise kept it from reaching the quantum ideal.

The researchers wrote that “our qubits coordinated better than any pre-written table allows.”

The experiment tested quantum contextuality, where measurement outcomes depend on what else is measured alongside them. On a contextuality measure where predetermined-outcome models score at most 4 and quantum theory allows 6, Willow registered 5.590 (with a reported uncertainty of 0.008).

Scaling up to 105 qubits

The larger runs are where the limits show.

A separate GHZ parity game asked up to 71 entangled qubits to deliver coordinated binary answers. The strongest classical winning rate approaches 50% as more players join, and Google stayed above that limit throughout its tests despite accumulating hardware noise. Researchers also verified genuine entanglement across 45 qubits.

The 105-qubit run tested a different, more limited challenge. In this hidden linear function task, a shallow quantum circuit had to produce a binary string satisfying mathematical constraints. The processor reached roughly 60% success, so its largest instance still failed in about four out of 10 runs.

Why Google says it matters

Google said the tasks exposed small, systematic errors that conventional randomized tests can hide. Its corporate blog called contextuality testing "a sensitive new method to catch subtle hardware errors," and Decrypt reported that it could potentially help engineers evaluate circuits used for chemistry and materials research. That's Google's claim about the method. The study doesn't present it as a finished tool.

The work follows Google's October report of a 40-fold error reduction in preparing magic states, a resource needed for fault-tolerant quantum computing.

For crypto readers, the background risk is the same as before. A future fault-tolerant quantum computer could break the digital signatures Bitcoin relies on, but Decrypt noted that it's still far from happening.

Frequently asked questions

What is the Mermin-Peres magic square game Google's Willow chip played?

Two players fill intersecting parts of a three-by-three grid without communicating. Their answers must obey conflicting row-and-column rules and match at the shared square. A prearranged classical strategy can succeed on no more than eight of nine assignments, or 88.9%.

Does Google's Willow result mean commercial quantum computing has arrived?

No. The 98.3% result came from four qubits, and the 105-qubit run tested a different, more limited task where the processor reached roughly 60% success. The result doesn't mean Google has a commercial quantum computer or beat every classical machine.

Does Google's quantum result threaten Bitcoin?

The study doesn't address Bitcoin. As background, the risk is that a future fault-tolerant quantum computer could break digital signatures, but Decrypt noted that remains far from happening.