Trending
Prev Story
Next Story

Google Quantum AI tests quantum contextuality on up to 105 qubits

The Next Web (TNW) •
November 19, 2026 • 08:06 PM
• By Ana Maria Constantin
5
5 reads 0 reactions

In a Science cover paper, Google Quantum AI shows contextuality at scale, winning a game 98.3% of the time against an 88.9% classical limit.

1 / 4
Swipe Photos

A smartphone displaying the Google search start screen.

Google Quantum AI says it has shown contextuality, a quantum property thought to power some algorithms, on up to 105 qubits, in a paper on the cover of Science. The tests also expose subtle hardware errors, but the paper does not claim an unconditional advantage over classical computers.

Google Quantum AI says a quantum property thought to power some algorithms still holds on up to 105 qubits. The research, published on Thursday on the cover of Science, tested the property on Google’s superconducting quantum processors.

The property is called contextuality. It means a measurement’s result is not fixed in advance, but depends on which other measurements are made alongside it. Until now, it had mostly been shown in systems of just a few particles.

In one test, Google’s qubits won a two-player game, the Mermin-Peres magic square, 98.3% of the time. The best possible classical strategy wins only 88.9% of the time. A larger task using 105 qubits also beat classical strategies with limited resources.

Google says the tests double as a sensitive way to catch subtle hardware errors that other benchmarks miss. It says this could help pave the way for uses in chemistry and materials science, though the experiments did not simulate either.

The paper is careful about its limits. It says there is no unconditional proof that quantum computers beat classical ones in general, and its quantum edge in one game shrank as more qubits joined. The work first appeared as a preprint in December 2025.

Error-free hardware remains the industry’s big goal. Google has said it wants a useful, error-corrected quantum computer by 2029, and Microsoft, which unveiled its Majorana 2 chip in June, has set the same target.

IBM agreed to buy HRL as it showed a chip that corrects its own errors , and Nvidia has simulated a 104-qubit error-corrected machine. Investors are piling in too, with Quantinuum raising $1.68B in its IPO in June.

Get the most important tech news in your inbox each week.

Prev Swipe Story Next

Related Stories

Technology
Build #2612