Proposed fault-tolerant throughput metric that scores logical operations per second with the classical decoder's accuracy, throughput, and latency folded in.
QLOPS (Quantum Logical Operations Per Second) is a proposed metric for benchmarking fault-tolerant quantum computers, introduced in 2025 by Linghang Kong and Fang Zhang (Zhongguancun Laboratory) and Jianxin Chen (Tsinghua University) in arXiv:2507.12024. It scores a machine’s logical throughput (logical qubits times the rate of logical operations per second) with the classical decoding stack’s accuracy, throughput, and latency folded in, factors the authors fault Microsoft’s rQOPS for ignoring. A whole-machine throughput score in spirit, it sits here alongside the fault-tolerant error-correction metrics it builds on.
How it works
QLOPS comes from modeling the full hardware-code-decoder stack, not from running a circuit protocol. For a code block encoding k logical qubits at distance d, with syndrome-extraction cycle time t_SEC and decoder reaction time t_r, a logical operation takes about ceil(t_r/t_SEC) + d cycles, giving Q = k / ((ceil(t_r/t_SEC) + d) · t_SEC) logical operations per second. Decoder accuracy enters through the achievable distance and logical error rate; code rate enters through k. The paper evaluates the metric on a Willow-like superconducting surface-code stack and a neutral-atom machine running generalized bicycle codes, grounded in a resource analysis of factoring RSA-2048.
Strengths and limitations
QLOPS is the first named fault-tolerant throughput metric to price in the classical decoder pipeline and code rate, and unlike rQOPS it has a peer-reviewed definition (ACM Transactions on Quantum Computing 7(2), article 14, 2026). So far, though, it is a modeling exercise: scores are computed for hypothesized stacks rather than measured by a vendor-run protocol, and results inherit the architecture model’s assumptions. It is easily confused with rQOPS (2023), which counts reliable operations at a target logical error rate without modeling the decoder; it complements memory-level figures like the logical error rate per round rather than replacing them.
Notable results
A maintained proposal, not yet an adopted standard: peer-reviewed in ACM Transactions on Quantum Computing (April 2026; arXiv v2, April 22, 2026), but as of July 2026 no hardware vendor reports QLOPS figures. IBM’s measured CLOPS remains the nearest NISQ-era analogue.