Thousands of circuits run on Quantinuum’s trapped-ion H2 systems produced scores above the best possible classical limit, with the largest experiment using 55 qubits on a 37-bit complement-sampling game.
The Nature Communications study says that gap grows exponentially and can be verified efficiently on a classical computer, avoiding the unproven computational-hardness assumptions used in many earlier quantum-advantage claims.
For the 37-bit instance, the theoretical separation between ideal quantum and best classical strategies exceeded 137 billion to one, though noise caused the biggest deviation from ideal performance in that largest test.
The result still depends on trusted components: the referee’s state preparation must be assumed reliable, and putting referee and player on the same processor means the setup is not yet a fully independent two-machine test.
Researchers said modest hardware gains could extend the method, but much larger demonstrations will likely require fault-tolerant error correction, and the game does not by itself show a general-purpose or commercial quantum advantage.
As quantum systems outpace classical verification, could this new complement sampling game be the ultimate key to trusting beyond-classical computation?
A quantum computer finally beat a mathematically proven classical limit, but will hardware noise eventually erase this exponential advantage before commercialization?
If one quantum machine can simulate both referee and player, what happens when two distinct computers are linked by a real quantum channel?