Google Willow chip wins 98.3% of a logic game classical tops at 88.9%
Google's Willow quantum chip won 98.3% of rounds of a logic puzzle where any classical strategy tops out at 88.9%. It is a test of quantum contextuality.

Google's Willow quantum chip just beat a logic puzzle that no classical strategy can fully win. It took 98.3% of the games. The best a classical player can do is 88.9%.
The result comes from Google Quantum AI and is published in the journal Science. It is about quantum contextuality, a property less famous than entanglement, though the experiment uses entangled qubits too. Michel Devoret, the French physicist who won the 2025 Nobel Prize in physics, is among the co-authors. The team had posted the work on arXiv in December 2025.
What is quantum contextuality?
In simple terms, the result of a measurement depends on the other compatible measurements made alongside it, its "context". So you can't assume the values were fixed in advance. Einstein supposedly asked whether the Moon is still there when nobody looks, defending the idea that properties exist before they are measured.
Google says it measured contextuality on Willow with up to 105 qubits. Until now it had mostly been seen on small systems with a handful of particles.
How does the magic square game work?
The game uses a 3 by 3 grid to fill with +1 and −1, with each row multiplying to +1 and each column to −1. That is impossible. Alice gets a row, Bob gets a column, and without talking they fill them in. They win if they write the same number in the shared cell.
Any strategy agreed in advance amounts to a fixed grid, which fails on at least one of the nine row-column pairs. That caps it at 8 wins out of 9, or 88.9%. With two pairs of entangled qubits, theory says you can win every time.
Google played each of the nine versions 100,000 times on Willow and won 98.3% on average, about one loss in 60. The team blames noise in the chip.
A test that doesn't move
Google also ran a Bell-Kochen-Specker test, which boils correlations down to one score. Pre-written results can't exceed 4, quantum mechanics allows up to 6, and Willow scored 5.59. Because that ceiling is set by physics, not by what supercomputers can do, it can't be caught up by faster classical machines. Google says it also revealed subtle drifts in a qubit's state that random-circuit tests tend to average away.
The larger-scale tests, a GHZ game up to 71 qubits and a hidden linear function problem up to 105, show a less clear edge. Google admits there is no immediate application.
Apps worth a look
Tap the right tiles, add them up, hit the target. Easy to learn, hard to master. Free, for all ages.
Talk to an AI conversation partner; every sentence comes back corrected and explained in your language. 9 languages, A1 to C2.
Open EDIFACT and X12 messages, see every segment explained, convert to JSON/XML, export PDF and Excel. Runs locally, no upload.
Pick who it is for, the occasion and their personality. The search engine reads product descriptions from local shops and ranks what fits. Free, no sign-up.
More from this day
See all →
Philips P79: the cardboard 'app' for a 1980s vacuum still works

GTA San Andreas house in GTA 6: where to find it in Leonida

TikTok AI video checker: upload any file, but it has a catch

Kakao founder Kim Beom-su acquitted again in SM stock case
More in: Science
See all →
AT 2025abao: astronomers watch one star swallow another

KT and LG send an image over quantum signals with QSDC

Walk score and steps: moving to a walkable city adds 1,100 a day
