New benchmark finds quantum computers 100,000x short of scientific utility

Current quantum computers remain roughly five orders of magnitude below the computational capability needed to tackle recognized scientific problems, according to a new benchmarking study posted on the arXiv preprint server. The finding, from researchers at Sandia National Laboratories with contributions from Quantinuum and NVIDIA ↗ 2.78%, introduces a framework called QUOPS — the Quantum Universal Operations Performance System — that measures both the scale of circuits a machine can execute and the speed at which it completes them.thequantuminsider+1

A New Yardstick for Quantum Machines

Quantum hardware makers typically describe their systems using qubit counts, gate accuracy and gate speed, but those component-level metrics do not always predict how a complete machine will perform on a substantial calculation. QUOPS instead treats the quantum computer as a whole system, feeding it randomized circuits of varying widths and depths, then checking whether the output passes a predefined accuracy threshold.thequantuminsider+1

The benchmark distills performance into two numbers: Q, representing the largest qualifying circuit the machine can run, and omega, the effective number of quantum operations completed per second at that scale. Together, they capture a trade-off that single specifications obscure.thequantuminsider

Trade-Offs Across Architectures

Researchers tested the benchmark on processors from Google Alphabet Inc. ↗ 0.78%, IBM International Business Machines Corporation ↘ 0.85% and Quantinuum. Google's Willow and IBM's Boston, both superconducting systems, ran operations faster and produced higher QUOPS rates. Quantinuum's Helios trapped-ion processor, which provides all-to-all qubit connectivity, successfully handled larger circuits and achieved a higher Q score but operated at a lower rate.quantinuum+1

The comparison does not crown one architecture as superior. Rather, the results show that machines reach performance through different combinations of accuracy, speed and connectivity — trade-offs that may matter differently depending on the workload.thequantuminsider

The Road to Scientific Utility

By mapping resource requirements for problems such as factoring RSA-2048 and calculating FeMoco energy eigenvalues, the researchers estimated target Q scores on the order of hundreds of millions. Current hardware falls about 100,000 times short of those targets, lending support to the ongoing push toward fault-tolerant quantum computing, which uses error correction to protect calculations from accumulated noise.quantumcomputingreport+1

The team also applied QUOPS to a small fault-tolerant setup using up to eight logical qubits on Quantinuum's Helios-1, demonstrating that the same benchmark can track both present-day physical-qubit machines and emerging logical-qubit systems. The paper has not yet undergone peer review.thequantuminsider+1