Friday, 20 November 2026NewsWorldBusinessTech
Latest

Google Quantum AI Beats Classical Limit With Willow Processor, Study Finds

Google Quantum AI’s Willow processor won a coordination game 98.3 percent of the time on Thursday beating the 88.89 percent classical limit.

A new study in Science shows that quantum contextuality survives at scale on superconducting processors. This property allows quantum systems to outperform classical strategies in tasks involving up to 105 qubits.

Researchers used the “Mermin-Peres magic square game” to demonstrate this effect. In this puzzle, two players fill a 3-by-3 grid with +1 or -1. Every row must multiply to +1, but every column must multiply to -1.

No fixed grid can satisfy these contradictory rules. A classical strategy can win at most eight of the nine possible rounds, according to Google’s official announcement. This occurs because row conditions require an even number of -1 entries. Column conditions require an odd number.

The Willow processor achieved a 98.3 percent success rate in this game. Crypto Briefing reported that nine assignments were each repeated 100,000 times. In this four-qubit experiment, players shared two entangled pairs and received two qubits each.

Hardware noise caused about one loss in every sixty rounds. Google attributed the remaining losses to these chip imperfections rather than flaws in the quantum strategy.

Google quantum contextuality willow chip
Photo: XenoSpectrum

Researchers scale contextuality to 105 qubits

Earlier demonstrations of contextuality were limited to small systems of a few particles. These new experiments extended tests to much larger arrays. The team implemented an N-player “GHZ parity game”.

In this game, each player corresponds to a qubit and the group shares an entangled GHZ state. Players return a one-bit answer without communicating. They win if the parity of the sum of their answers matches the parity of half the sum of the input bits.

The win rate was 75 percent at N = 3. This rate decayed toward 50 percent as more players were added.

The researchers also used constant-depth quantum circuits to tackle the “Hidden Linear Function (HLF) problem”. The target was a symmetric binary matrix representing a quadratic form modulo 4. These experiments outperformed constrained classical strategies in tasks spanning up to 105 qubits, Quantum Computing Report said.

Demonstrating quantum contextuality at scale on the Willow processor — Google Quantum AI

The success probability fell from about 1 to about 0.6 at full scale. The system maintained an effective depth of roughly 4 to 6 layers. This work builds on a 2018 Science paper by Bravyi, Gosset, and König.

That paper proved that constant-depth quantum circuits can solve certain HLF problems more efficiently than classical circuits. Those classical circuits have specific constraints on depth and fan-in.

Task Qubit Scale Reported Performance
Mermin-Peres magic square game 4 qubits (2 Bell pairs) 98.30% ± 0.01% win rate
GHZ parity game Up to 71 qubits Exceeded classical win rate across all sizes
2D Hidden Linear Function (HLF) 9 to 105 qubits Success probability > 0.5

Tests provide new benchmarks for hardware

Structured parity and hidden-function tests require exact multi-qubit phase agreement. These tests can identify small coherent over-rotations. The researchers converted the fraction of correct outputs into an effective depth.

Effective depth is a measure of time-to-solution. It reflects the extra runs required to reach a correct answer.

Google quantum ai demonstrates contextuality across qubits
Photo: Unite.AI

The study also used a state-independent Kochen-Specker-Bell test to quantify contextuality. This test used 540 sequential QND measurements repeated 1,000 times. It measured the machine rather than a particular input.

The processor recorded a value of 5.590 or 5.618 ± 0.005. This exceeds the classical limit of 4 and approaches the quantum limit of 6, Interesting Engineering said. A noncontextual model is constrained to χKSB ≤ 4.

Contextuality is a potential “magic ingredient” for quantum computing, researchers said. While entanglement often receives attention, theory suggests contextuality may be the deeper resource that powers certain quantum algorithms, Google Research said.

The findings also connect contextuality with magic states. These states are essential for running full algorithms on error-corrected quantum computers.

Library google s willow quantum chip wins of magic square g
Photo: Crypto Briefing

The Willow processor is a 105-qubit superconducting transmon chip. It has been Google’s flagship since its December 2024 announcement. In October 2025, Google ran its Quantum Echoes algorithm on the chip.

That run claimed a verifiable quantum advantage 13,000 times faster than classical hardware. These current experiments did not perform practical molecular simulations. However, the ability to detect coherent errors helps diagnostic options for future work in chemistry and materials science.

This progress follows a long-term effort by Google Quantum AI to build large-scale, error-corrected quantum computers. The research division was launched in 2012, building on an initial concept from 2006 intended to accelerate machine learning. The group has previously reached major milestones, including a 2019 claim of quantum supremacy using a 53-qubit Sycamore processor and a 2023 demonstration of a logical qubit prototype.

The underlying research was first submitted as an arXiv preprint on December 1, 2025.