New benchmark, QuSquare, tests near-term quantum devices

Researchers from the Department of Physical Chemistry, University of the Basque Country UPV/EHU and BCAM, Basque Center for Applied Mathematics in Bilbao, Spain have published a new benchmark suite called “QuSquare” designed to evaluate current quantum devices. Published in Quantum Science and Technology, Number 3, with DOI 10.1088/2058-9565/ae917c, QuSquare focuses on assessing the performance of quantum computers before they reach fault tolerance. The work addresses a key challenge in quantum computing: achieving fair, scalable, and consistent evaluations of these pre-fault-tolerant systems.

QuSquare: Benchmarking for Pre-Fault-Tolerant Quantum Devices

Unlike many existing efforts focused on the distant promise of fault-tolerant quantum computation, QuSquare specifically targets pre-fault-tolerant devices, acknowledging the current limitations of near-term quantum technology. This focus allows for a more practical assessment of quantum processors and provides a pathway for tracking progress as the field matures. The developers detail their work in Quantum Science and Technology, outlining a suite designed to evaluate both system- and application-level performance.

The creation of QuSquare responds to a critical gap in the field; the proliferation of diverse hardware architectures demands fair, scalable, and consistent evaluation methods. The researchers note that without rigorous benchmarks, there is a risk of misleading performance metrics that may distort research priorities, potentially hindering genuine advancement.

QuSquare aims to mitigate this risk by incorporating quality attributes such as relevance, reproducibility, fairness, verifiability, and scalability into its design. The suite consists of four distinct benchmark tests, each probing different aspects of quantum hardware capability, from implementing Clifford gates to simulating quantum many-body dynamics. Specifically, the QuSquare suite includes a Partial Clifford randomized benchmark assessing gate accuracy, a multipartite entanglement test evaluating the generation of complex quantum states, a transverse field Ising model (TFIM) Hamiltonian simulation benchmark, and a data re-uploading quantum neural network (QNN) benchmark.

These tests were chosen to provide a comprehensive evaluation, moving beyond component-level assessments to examine performance at the system and application levels. To ensure transparency and facilitate wider adoption, the team has also provided an open-source implementation of the QuSquare benchmark suite, allowing other researchers to readily utilize and contribute to its development.

The design principles behind QuSquare prioritize evaluating the quality of quantum computers by assessing noise in circuit implementation and the ability to execute specific applications. The researchers emphasize that the suite is not simply about achieving high scores on arbitrary tasks, but rather about providing meaningful insights into the strengths and weaknesses of different quantum platforms. They explain this holistic approach to its development.

The team intends for QuSquare to contribute to the development of future benchmarking standards and enable more informed comparisons between emerging quantum technologies. It was received on February 16, 2026, revised on June 25, 2026, and accepted on July 28, 2026.

QuSquare Protocols: Relevance, Reproducibility, and Fairness

This focus on current, imperfect hardware distinguishes QuSquare from much of the existing discussion surrounding quantum computing, which often centers on the anticipated capabilities of fault-tolerant machines. The team’s work addresses a critical need for consistent and comparable performance metrics as the number of diverse quantum architectures continues to grow rapidly. The impetus behind QuSquare stems from the recognition that a lack of standardized benchmarks risks creating a skewed understanding of progress in the field.

To counter this, QuSquare prioritizes quality attributes like relevance, reproducibility, fairness, verifiability, and scalability in its design, aiming to provide a more holistic and reliable assessment of quantum hardware. These tests were selected to provide a comprehensive evaluation, encompassing both fundamental quantum operations and the execution of more complex algorithms.

The researchers from the Department of Physical Chemistry, University of the Basque Country UPV/EHU and BCAM, Basque Center for Applied Mathematics detail that the suite is composed of both system- and application-level tests to assess the performance of pre-fault-tolerant quantum devices, ensuring it satisfies scalability, fairness, relevance, verifiability, and reproducibility. The data supporting the findings of this study are openly available, further promoting collaboration and validation within the quantum computing community.

Partial Clifford Randomized Benchmark for System Evaluation

Central to the QuSquare suite is a benchmark, a test designed to assess the accuracy of quantum processors in implementing a specific subset of quantum gates. These Clifford gates are particularly relevant because they form the foundation for many quantum error correction schemes, even in imperfect, near-term hardware. The benchmark doesn’t simply measure success or failure, but quantifies the fidelity with which these gates are executed, providing a nuanced understanding of a device’s strengths and weaknesses.

Multipartite Entanglement Test within QuSquare Suite

The ability of a quantum processor to generate complex entangled states is a critical measure of its potential, and the QuSquare suite directly assesses this capability with a dedicated multipartite entanglement test. This benchmark moves beyond simply creating entanglement between two qubits, instead evaluating the fidelity with which a quantum device can produce genuine multi-particle entanglement, a resource essential for advanced quantum algorithms and communication protocols.

The multipartite entanglement benchmark within QuSquare focuses on quantifying the quality of these entangled states, not just their existence. This is achieved by carefully characterizing the resulting state using a series of measurements and comparing the observed correlations with the theoretical predictions for a perfect GHZ state. The benchmark’s design prioritizes a clear determination of whether the observed entanglement is genuinely multipartite, distinguishing it from correlations that could be explained by simpler, separable states.

Crucially, the QuSquare approach to evaluating multipartite entanglement incorporates strategies to mitigate the impact of hardware-specific limitations. The researchers from the Department of Physical Chemistry, University of the Basque Country UPV/EHU and BCAM, Basque Center for Applied Mathematics considered the challenges posed by qubit connectivity, gate errors, and measurement imperfections when designing the benchmark protocol. Parameters within the test are adjustable, allowing researchers to tailor the complexity of the generated GHZ state to the capabilities of the specific quantum device under evaluation.

This adaptability is key to ensuring a fair comparison across diverse hardware architectures. The researchers explain that the benchmark’s metrics are designed to provide a quantitative assessment of entanglement quality, allowing for a precise comparison of performance across different quantum computing technologies.

The resulting data from the multipartite entanglement test is then used to calculate a fidelity score, representing the degree to which the generated state matches the ideal GHZ state. This score provides a clear and concise measure of the quantum processor’s ability to create and maintain complex entanglement.

The QuSquare suite’s open-source implementation facilitates reproducibility, allowing other research groups to independently verify the results and contribute to the development of standardized benchmarking methodologies. This detailed assessment of multipartite entanglement generation is a vital component of the QuSquare suite, offering a rigorous and scalable method for evaluating the performance of pre-fault-tolerant quantum devices and guiding future hardware development.

Transverse Field Ising Model Simulation in QuSquare

This benchmark moves beyond simple gate fidelity tests by evaluating a quantum computer’s performance on a problem directly relevant to materials science and fundamental physics. The TFIM benchmark specifically assesses a quantum processor’s capacity to simulate the time evolution of a spin system under the influence of both a transverse magnetic field and interactions between neighboring spins.

The selection of parameters for the TFIM simulation is not arbitrary; the researchers emphasize the importance of choosing values that are both computationally challenging and physically relevant. By simulating systems with varying degrees of complexity, they can assess the scalability of the quantum hardware and identify potential bottlenecks in performance.

Data Re-uploading Quantum Neural Network Benchmark

The QuSquare benchmark suite, detailed in Quantum Science and Technology, includes a dedicated assessment of data re-uploading quantum neural networks, a method gaining traction for its potential to mitigate the impact of noise on near-term quantum devices. This benchmark specifically evaluates a quantum processor’s ability to perform calculations central to machine learning tasks, offering a performance gauge beyond standard gate fidelity measurements.

The data re-uploading quantum neural network benchmark within QuSquare focuses on classification problems, assessing the hardware’s performance as it processes and learns from input data. Unlike some quantum machine learning algorithms requiring extensive, deeply entangled states, data re-uploading techniques allow for shallower circuits, potentially reducing the impact of decoherence and gate errors common in current quantum computers.

The QuSquare implementation allows users to select parameters influencing the complexity of the neural network and the dataset used for training, enabling evaluation across a range of problem sizes and hardware capabilities. This flexibility is crucial for determining the scalability of quantum machine learning algorithms on different platforms. Crucially, the QuSquare approach incorporates specific metrics to quantify performance beyond simple accuracy scores.

The benchmark evaluates the ability of the quantum processor to maintain signal fidelity throughout the repeated data re-uploading cycles, a critical factor in determining the quality of the learned model. Performance is assessed by analyzing the resulting classification accuracy and correlating it with metrics that measure the entanglement generated and preserved during the computation.

The researchers from the Department of Physical Chemistry, University of the Basque Country UPV/EHU and BCAM, Basque Center for Applied Mathematics emphasize that these metrics provide a more nuanced understanding of hardware limitations than traditional benchmarks, pinpointing specific areas where improvements are needed to support quantum machine learning applications. The data supporting these findings are openly available, facilitating independent verification and further research.

QuSquare’s System and Application-Level Performance Tests

The suite’s creation responds to a growing need for standardized evaluation methodologies as quantum technologies diversify, with numerous hardware architectures emerging, each possessing unique strengths and limitations. The Partial Clifford randomized benchmark assesses the accuracy of implementing a subset of Clifford gates, crucial for quantum error correction protocols, while the multipartite entanglement benchmark evaluates a quantum processor’s ability to generate complex entangled states, a key resource for quantum computation.

These principles guide the suite’s ability to assess increasingly large devices without favoring specific technologies and ensure that evaluations are based on clear, repeatable protocols. The team from the Department of Physical Chemistry, University of the Basque Country UPV/EHU and BCAM, Basque Center for Applied Mathematics has also made the open-source implementation of QuSquare publicly available to encourage adoption and independent verification of results. This nuanced evaluation provides a more detailed understanding of hardware limitations than traditional metrics alone.