QUOPS moves beyond qubit counts to measure quantum computers as complete systems

A quantum computer’s qubit count and individual gate benchmarks do not, on their own, show what it can accomplish on a real workload. QUOPS—short for Quantum Universal Operations Performance System—takes a different approach: it measures how large a broadly useful quantum computation a system can complete reliably, and how fast it completes that work.

The work was led by Sandia National Laboratories, with Quantinuum contributing hardware demonstrations and NVIDIA providing an open-source CUDA-Q implementation. 13

What QUOPS measures

QUOPS is an architecture-independent benchmark intended for both physical and logical qubits. Rather than testing one specific algorithm, it runs the same randomized, computationally universal circuits across systems with different shapes and sizes. Each workload must clear a fidelity- or polarization-based success threshold to count as a successful execution. 23

That produces two complementary measurements:

• Q, the QUOPS score: the size, measured in quantum operations, of the largest circuit that reliably passes the benchmark.
• ω, the QUOPS rate: the effective rate at which those successful operations are performed, expressed in QUOPS per second.

The distinction matters. One machine may execute smaller workloads rapidly, while another may reliably sustain substantially larger circuits at a lower rate. QUOPS is designed to show both sides of that trade-off. 12

A benchmark for the entire system

Running a quantum circuit depends on more than the processor’s headline specifications. Circuit compilation and mapping, qubit routing, hardware connectivity, control systems, measurement and noise can all affect whether a computation succeeds.

QUOPS is intended to bring those costs into a single system-level result, rather than leaving them as disconnected hardware specifications. 12 In other words, it is not a ranking based on one headline number; it is an attempt to map the boundary of a complete quantum system’s reliable computing capability. 3

Willow, Boston and Helios: speed versus circuit scale

The first comparisons showed a clear trade-off between circuit size and execution speed:

• Google’s superconducting Willow processor and IBM’s superconducting Boston processor achieved higher operation rates.
• Quantinuum’s trapped-ion Helios supported larger circuits that met the benchmark’s success criterion. 45

The proposed explanation is that fast superconducting gates help boost execution rate, while Helios’s high accuracy and effectively all-to-all connectivity reduce the routing overhead needed to link qubits. That can limit the accumulation of errors before a circuit becomes too large to run reliably. 25

The reported physical-qubit Q scores were 1,504 for Helios-1, 216 for Willow and 204 for IBM Boston. Those figures do not establish one universal winner. They describe where each system sits on a particular balance between reliable circuit scale and throughput. 7

How far current systems are from large-scale applications

The preprint estimates targets of roughly 2.5 × 10⁸ QUOPS for factoring RSA-2048 and 3.4 × 10⁸ QUOPS for high-accuracy FeMoco calculations. On that basis, current physical-qubit systems remain about five orders of magnitude—or roughly 100,000 times—short of the scale estimated for these demanding workloads. 57

These are resource estimates, not a timetable. They depend on assumptions about algorithms, error-correction codes and hardware architectures, so they do not predict when a particular machine will factor RSA-2048 or solve a FeMoco problem.

Extending QUOPS to logical qubits and fault tolerance

The benchmark was also used in a fault-tolerant demonstration involving up to eight logical qubits on Helios-1. Sandia says this marked the first benchmarking of a fault-tolerant processor. 12

That matters because QUOPS is not limited to today’s noisy physical qubits. Its basic approach can also assess systems using encoded logical qubits and error correction. The demonstration does not mean fault-tolerant hardware already outperforms physical-qubit systems by every measure. Its significance is that it provides a shared way to track progress toward larger fault-tolerant systems. 13

How to interpret the results

QUOPS is a useful effort to shift the question from “How many qubits does it have?” to “How much reliable, universal quantum work can it actually complete?” The early tests suggest that connectivity and accuracy can increase the maximum feasible circuit size, while rapid gates can raise the rate of execution.

Still, the findings come from an arXiv preprint that has not yet been peer reviewed. The benchmark’s definitions, success thresholds and extrapolations to large-scale applications should therefore be treated as preliminary until they have been independently examined and reproduced. 5