Google’s Willow quantum chip wins 98.3% of magic square games, beating the classical limit Google Quantum AI's 105-qubit Willow superconducting transmon processor achieved a 98.3% success rate in the four-qubit Mermin-Peres magic square game, exceeding the 88.9% classical ceiling, according to a peer-reviewed study published November 19, 2026, on the cover of Science. The team ran nine game assignments 100,000 times each and extended contextuality tests to multi-player GHZ-type games and hidden linear function problems across all 105 qubits; most losses traced to residual hardware noise rather than the quantum strategy. The result, first posted as an arXiv preprint in December 2025, demonstrates quantum contextuality at scale and positions the magic square game as a hardware diagnostic with a known answer key. Google’s Willow quantum chip wins 98.3% of magic square games, beating the classical limit Photo: Jakub Pabis / Pexels A peer-reviewed study in Science shows the 105-qubit processor demonstrating quantum contextuality from four qubits up to 105 Google https://cryptobriefing.com/markets/alphabet/ Quantum AI’s Willow processor just won a game that classical computers mathematically cannot win as often. In the four-qubit version of the Mermin-Peres magic square game, Willow posted a 98.3% success rate , clearing the classical ceiling of 88.9%. That gap is the point. A quantum chip beating a classical strategy at a puzzle sounds like a party trick, but the result is a clean, measurable demonstration of one of quantum mechanics’ strangest properties. What Willow actually did The findings were published on November 19, 2026, in the journal Science , where the work landed on the cover. A preprint version first appeared on arXiv in December 2025, so the peer-reviewed paper arrives roughly a year after the research community first got a look. Willow itself is a 105-qubit superconducting transmon processor. Google Quantum AI announced the chip in December 2024. The headline experiment centers on the Mermin-Peres magic square game. Think of it as a cooperative test where players cannot coordinate in real time, yet still need their answers to line up. Players relying on classical strategies hit a hard cap. Under the research findings, the best they can manage is 88.9%. Willow reached 98.3%. That is not a rounding-error improvement on the classical limit. It is a result classical logic says should be off the table. The testing was thorough. The team ran nine game assignments, executing each one 100,000 times. When Willow did lose, the culprit was mostly residual noise in the hardware. Put simply, the losses came from the chip’s imperfections, not from any flaw in the quantum strategy itself. AI, tech, and the markets they move—in one daily briefing. Daily. Free. Join 34,000+ readers across crypto, finance, and policy. Why contextuality matters The property on display here is called quantum contextuality. It means the outcome of measuring a quantum system depends on the experimental context in which you measure it. The team did not stop at four qubits. They extended the tests to multi-player GHZ-type games and hidden linear function problems, scaling up to all 105 qubits on the chip. The hidden linear function work relied on a shallow-circuit algorithm. A shallow circuit uses only a few layers of quantum operations, which matters because fewer steps give noise fewer chances to creep in and spoil the result. Running these tests across the full chip suggests the contextuality advantage is not confined to a small, carefully babysat corner of the hardware. It persisted as the system grew. A benchmark that doubles as a diagnostic One of the more practical takeaways is that the magic square game works as a hardware check. Because the classical limit is fixed and known, any shortfall from the ideal quantum score points to something going wrong on the chip. The research describes this testing as a way to surface subtle hardware errors. That turns an abstract physics demonstration into something closer to a stress test with a known answer key. Since most of Willow’s losses traced back to residual noise, the game effectively maps where the remaining imperfections live. The broader quantum picture Willow has been Google’s flagship quantum chip since its December 2024 announcement. This latest paper adds a fresh line to its resume, focused less on raw speed and more on proving that the device behaves in genuinely non-classical ways. The publication timeline is also worth noting. Moving from an arXiv preprint in December 2025 to the cover of Science in November 2026 means the results have now passed formal peer review. What this means For the quantum computing industry, the result is a credibility marker. Demonstrating contextuality at 98.3% against an 88.9% classical ceiling gives Google a hard number to point to when it argues its hardware does something classical machines cannot. The scale also matters. Showing quantum behavior across up to 105 qubits, rather than a handful, is the kind of evidence rival labs will likely be measured against. The research itself acknowledges that practical applications have been slow to roll out, so a winning streak at a logic game is not the same as commercially useful computation. The more immediate thing to watch is noise. Willow’s losses came largely from residual hardware noise, and how quickly Google and its competitors drive that down will determine whether demonstrations like this one turn into machines that solve problems people will pay for. Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy https://cryptobriefing.com/editorial-policy/ .