Researchers have simultaneously benchmarked nine commercially accessible quantum processors from five vendors, revealing significant performance differences across superconducting and trapped-ion technologies. The work estimates the fidelity of Greenberger-Horne-Zeilinger (GHZ) states, a complex form of quantum entanglement, to assess processor capabilities using a benchmark from the QuSquare suite. Results show the best superconducting processors reached 27 qubits, less than a quarter of their total register size, while a 12-qubit trapped-ion processor successfully completed the benchmark with its full register. These findings provide a baseline for future cross-platform comparisons of quantum processors.

Cross-Platform Benchmarking with GHZ States Assesses Quantum Processors

Statistical decision rules significantly impacted benchmark outcomes, reducing the number of processors successfully passing tests from nine to four when requiring certified genuine multipartite entanglement (GME). This finding highlights the sensitivity of entanglement benchmarks to evaluation criteria and emphasises the need for standardized protocols when comparing quantum processors. The work assessed performance by determining the largest GHZ state a processor could reliably generate, a metric directly tied to its ability to sustain complex quantum correlations.

Median calibration data proved unreliable predictors of benchmark success, suggesting that processor specifications alone do not fully capture performance capabilities. Single-run experiments revealed statistically significant, non-monotonic features, indicating run-to-run variability in qubit mapping and calibration beyond typical shot noise. This variability suggests that achieving consistent performance requires careful control of experimental parameters and robust calibration procedures.

The study’s methodology involved remote cloud access to nine commercially available quantum processors, allowing for direct comparison across different hardware platforms without the need for on-site experimentation. This approach reflects the performance experienced by end users, providing a realistic assessment of processor capabilities in a practical setting. GHZ states are valuable resources for quantum metrology, fault-tolerant protocols and multiparty communication, making their reliable generation a key indicator of quantum computational power.

The benchmark employed direct fidelity estimation to assess the quality of the generated GHZ states, searching for the largest state whose fidelity met the threshold for genuine multipartite entanglement. “As quantum computing hardware continues to evolve, benchmarking has become essential not only for tracking technological progress and comparing the performance of quantum processors, but also for enabling researchers and organizations to make informed decisions when selecting the most appropriate quantum computing platform for their applications,” the paper states.

Researchers found that hardware specifications, such as qubit count, coherence times and gate fidelities, offer only a partial characterization of a quantum processor’s capabilities. Performance metrics that capture common capabilities across platforms are particularly valuable for cross-platform benchmarking, and the generation of genuine multipartite entanglement is one such capability.

Nine Processors Compared: Superconducting and Trapped-Ion Technologies

Performance discrepancies emerged sharply when evaluating the largest Greenberger-Horne-Zeilinger (GHZ) state each processor could reliably generate, a key indicator of entanglement quality; while some superconducting processors reached a capacity of 27 qubits under the benchmark’s criteria, the 12-qubit trapped-ion processor passed with its full register. Researchers acknowledge the challenges of fairly implementing the benchmark across such diverse hardware and access methods, and are actively working to expand the trapped-ion side of the comparison. The work also notes that while adding more sophisticated error mitigation techniques could increase the metric’s value, it would likely not alter the relative performance rankings between the processors.

QuSquare Suite Measures Fidelity of GHZ-State Generation

Direct Fidelity Estimation, a method independent of qubit count for stabilizer states, underpinned the assessment of Greenberger-Horne-Zeilinger (GHZ) state fidelity across nine processors accessed remotely via cloud services. The study deliberately employed each provider’s default compilation settings and refrained from error mitigation techniques, mirroring the experience of end users accessing these systems.

Detailed analysis of individual fidelity estimates revealed stability in the maximum GHZ-state size despite variations between executions, with further investigation detailed in an appendix. Researchers stress that $N^{*}$ represents the largest GHZ size compatible with genuine multipartite entanglement, not necessarily the largest for which it is certified, and that the decision rule employed influences the results.

The work highlights a potential asymmetry introduced by the toolchains used in the benchmarking process, suggesting that variations in compilation and control software could contribute to performance differences beyond the inherent capabilities of the quantum hardware itself. The team’s findings demonstrate the importance of evaluating quantum systems not just on theoretical qubit counts, but on their ability to reliably generate and maintain entanglement in complex states like GHZ states, a key step towards realizing the potential of quantum computation for practical applications. The complete dataset supporting these results is openly available at Zenodo under a CC BY 4. 0 license, including the ideal and executed circuits and the calibration snapshots used during the experiments.

Benchmark Size Discrepancy: Physical Qubits vs. GHZ State Scale

Experiments performed via remote cloud services revealed a significant disparity between total qubit count and successful multipartite entanglement generation, with the largest GHZ states achieved on tested processors comprising less than a quarter of their physical qubit registers. This suggests that simply increasing qubit numbers does not guarantee improved performance in complex quantum tasks.

The study’s methodology deliberately excluded expert-level optimization techniques, such as device-specific compilation or advanced error mitigation, to reflect the experience of typical cloud users. This contrasts with a recent demonstration of a 120-qubit GHZ state on an IBM Heron processor, which relied on post-selection, dynamical decoupling and readout-error mitigation, procedures not standard in default cloud workflows.

Analysis of fidelity estimates revealed fluctuations attributable to compiler-selected physical qubits and calibration drift between experimental runs, rather than random noise, highlighting the influence of software-level factors on hardware performance. The ability to generate genuine multipartite entanglement, as demonstrated by the GHZ states, is important for diverse applications in quantum information science, including quantum communication, quantum sensing, and distributed quantum computing.

Calibration Data Shows Limited Predictive Power for Benchmarking

While readily available calibration metrics like single- and two-qubit error rates are routinely used to assess hardware quality, the research demonstrates these figures poorly predict success in creating Greenberger-Horne-Zeilinger (GHZ) states, a benchmark of multipartite entanglement. The team found that on superconducting processors, the error rate per qubit during GHZ state generation exceeded the median two-qubit gate error by a factor of ten, highlighting the importance of factors beyond basic gate fidelity.

This scale of direct comparison is notable, moving beyond typical evaluations focused on single processors. Analysis of the results showed that the largest GHZ state successfully generated varied significantly between devices, and was not simply correlated with the best calibration numbers.

In fact, requiring certified genuine multipartite entanglement reduced the number of devices that succeeded at any tested size from nine to four. The study’s findings highlight that calibration data, while useful, provide an incomplete picture of a processor’s capabilities. The team identified readout errors, qubit idling times and crosstalk as significant contributors to overall GHZ state fidelity, factors not fully captured by standard calibration metrics.

“The outcome depends sensitively on the statistical decision rule,” the researchers write, emphasizing the impact of analytical choices on benchmarking results. This sensitivity suggests that standardized benchmarking protocols must account for both hardware characteristics and the specific software-level factors influencing performance.

Cloud Access Enables Remote Benchmarking of Nine Processors

The University of Deusto coordinated a study benchmarking quantum processors not within a single laboratory, but through remote cloud services, enabling simultaneous analysis of nine devices from five vendors. This approach bypassed the logistical hurdles of physically accessing and operating disparate quantum hardware, allowing for direct performance comparisons across technologies. The work focused on assessing each processor’s ability to generate Greenberger-Horne-Zeilinger (GHZ) states, a specific form of multipartite entanglement used as a benchmark for quantum capabilities.

The study’s findings highlighted the limitations of relying solely on calibration data to predict processor performance. These included the University of the Basque Country UPV/EHU, Ikerlan Technology Research Centre and Vicomtech Foundation.

👉 More information
🗞 Cross-Platform Benchmarking of GHZ-State Generation in Quantum Processors via Cloud Services
✍️ Sergio Hernández et al.
🧠 ArXiv: https://arxiv.org/abs/2611.14114v1

More like this