A research collaboration led by Sandia National Laboratories, in partnership with Quantinuum and NVIDIA, has introduced the Quantum Universal Operations Performance System (QUOPS). Detailed in a multi-institution paper published on arXiv (arXiv:2609.12146), QUOPS is an architecture-agnostic benchmarking framework designed to measure integrated quantum computer performance across both physical- and logical-qubit platforms, addressing limits in traditional component-level metrics such as raw qubit count or isolated gate fidelities.
QUOPS evaluates full system execution—incorporating compilation, error correction, syndrome decoding, and error mitigation—by executing randomized layers of arbitrary-angle single-qubit rotations, RP(θ), and CNOT gates across varying circuit widths (w) and depths. The benchmark yields two top-level system metrics: Q (QUOPS score), which defines the maximum circuit size s = 2 × w × depth successfully executed within a utility-motivated geometric volume (w² ≤ s ≤ w³) at a minimum mean process polarization threshold of 1/√e ≈ 61%; and Ω (QUOPS rate), which measures the net operational throughput in executed operations per second while accounting for sampling overheads from error mitigation or postselection.
| [ Experimental Cross-Platform QUOPS Performance Benchmarks ] | ||
|---|---|---|
| Hardware System & Modality | Architecture & Encoding Layer | QUOPS Score (Q) & Throughput (Ω) |
| Google Willow | • Physical Transmon (105 Qubits, 2D Grid) | • Q = 216 | Ω = 2.0 × 10⁷ QUOPS/s (Width w=6) |
| IBM ibm_boston | • Physical Transmon (156 Qubits, Heavy-Hex) | • Q = 204 | Ω = 3.1 × 10⁵ QUOPS/s (Width w=6) |
| Quantinuum H2-1 | • Physical Trapped-Ion (56 Qubits, QCCD) | • Q = 1,320 (1,392 w/ PS) | Ω = 353 QUOPS/s (Width w=12) |
| Quantinuum Helios-1 | • Physical Trapped-Ion (98 Qubits, QCCD) | • Q = 1,504 (1,824 w/ PS) | Ω = 303 QUOPS/s (Width w=16) |
| Quantinuum Helios-1 (FTQC) | • Logical Steane [[7,1,3]] Code (8 Logical Qubits) | • Q = 40 | Ω = 4.9 QUOPS/s (Width w=4) |
The publication provides direct cross-platform baseline measurements across leading hardware: Quantinuum’s Helios-1 achieved a physical score of Q = 1,504 (1,824 under leakage postselection), Google’s Willow achieved Q = 216 at 2.0 × 10⁷ QUOPS/s, and IBM’s ibm_boston reached Q = 204 at 3.1 × 10⁵ QUOPS/s. Additionally, the team executed QUOPS on a fully fault-tolerant logical architecture using up to 8 Steane-encoded [[7,1,3]] logical qubits on Helios-1, demonstrating logical Clifford+T circuit execution with active magic-state injection and syndrome extraction to achieve Q = 40 at Ω = 4.9 QUOPS/s.
By mapping resource requirements for canonical utility challenge problems into effective QUOPS circuit sizes—estimating targets of Q ≈ 2.5 × 10⁸ for factoring RSA-2048 and Q ≈ 3.4 × 10⁸ for FeMoco energy eigenvalue calculations—the framework quantifies the 5-order-of-magnitude gap remaining between contemporary hardware and utility-scale application requirements, establishing a standardized metric for government procurement, hardware roadmaps, and fault-tolerant architectural verification.
Review the technical overview on Quantinuum here, access the peer-reviewed preprint on arXiv here, and inspect open-source benchmark implementations on GitHub here.
September 14, 2026
