QUOPS Benchmark Maps a Five Order Gap to Quantum Utility
Sandia National Laboratories, Quantinuum, and NVIDIA published QUOPS, a cross platform benchmark showing today's best QPUs remain roughly five orders of magnitude short of recognized scientific utility targets.
What changed
Sandia National Laboratories, with Quantinuum and NVIDIA, introduced the Quantum Universal Operations Performance System, called QUOPS, in a 10 September 2026 arXiv preprint numbered 2609.12146. NVIDIA announced on 14 September 2026 that a QUOPS reference implementation is available inside its open source CUDA-Q platform.
QUOPS measures two linked quantities: Q, the size of the largest random circuit a machine executes above a fidelity threshold, and the execution rate in QUOPS per second. The authors applied the benchmark directly on physical qubits from Google Willow, IBM ibm boston, Quantinuum H2-1, and Quantinuum Helios-1.
Reported scores include Helios-1 at Q equals 1504 with 303 QUOPS per second, Google Willow at Q equals 216 with about 2.0 times ten to the seventh QUOPS per second, and IBM ibm boston at Q equals 204 with about 3.1 times ten to the fifth QUOPS per second. When the authors translate resource needs from recognized challenge problems into effective QUOPS circuit sizes, they conclude capability must grow by roughly five orders of magnitude, reinforcing the case for fault tolerant architectures.
Why it matters
Quantum procurement has lacked a hardware agnostic yardstick that pairs circuit depth with speed. Vendor qubit counts alone cannot tell R&D leaders whether budget should shift toward error correction, hybrid classical orchestration, or waiting for the next platform generation.
QUOPS also reports an early logical qubit run on Quantinuum Helios-1 using up to eight encoded logical qubits, giving capital committees a single framework to track fault tolerant progress instead of disconnected lab demos.
Who is affected
Quantum program managers at national labs, pharma computational chemistry leads, financial risk teams exploring quantum pilots, and investors comparing trapped ion versus superconducting roadmaps.
What to do next
Rebaseline any 2027 quantum pilot against QUOPS Q and rate targets from the preprint rather than raw qubit marketing sheets. If your vendor is not yet benchmarked, ask for a third party QUOPS run before renewing cloud credits.
What to watch
Whether IonQ and other trapped ion vendors publish QUOPS results at IEEE Quantum Week follow ups, and whether CUDA-Q Logical deployments change how quickly teams can simulate fault tolerant schedules before hardware arrives.
Sources
- Primary. arXiv, Benchmarking the computational power of quantum computers (10 September 2026). Defines QUOPS and reports experimental scores.
- Primary. Quantinuum, Introducing QUOPS (14 September 2026). Explains Q and rate metrics with platform comparison figures.
- Secondary. NVIDIA Newsroom, NVIDIA expands CUDA-Q for fault tolerant quantum computing (14 September 2026). Confirms QUOPS availability in CUDA-Q.