How quantum computers are compared: component tests of gates, system-level scores of whole processors and application suites that run small versions of real tasks. Each benchmark is described with what it measures, how, its limitations, who maintains it and whether it is European.
Benchmark results are mostly reported by the providers themselves; a higher score on one benchmark does not mean a better machine for every task.
The average error rate of a set of gates, under a general noise model including time- and gate-dependent errors.
Maintained by: A family of protocols from the research literature, implemented in most vendor toolkits.
European: No
How it works and its limitations Randomised benchmarking family
Local and global errors of whole cycles of multi-qubit operations, including cross-talk and correlated noise.
Maintained by: Proposed by Erhard et al. (2019).
European: Yes: First author and most co-authors are at the University of Innsbruck (Austria), with co-authors at the University of Waterloo and Quantum Benchmark Inc. (Canada) (Europe PMC record of the Nature Communications paper, read on 27 October 2026).
A single number for the largest random circuit of equal width and depth that a computer runs successfully.
Maintained by: Defined by Cross et al. (2019); results are mostly self-reported by vendors.
European: No
Speed: circuit layer operations per second, measured with an update of the quantum volume experiments.
Maintained by: Proposed by Wack et al. (2021).
European: No
How it works and its limitations CLOPS (circuit layer operations per second)
The fidelity of a connected layer of two-qubit gates across many qubits, and a size-independent error per layered gate derived from it.
Maintained by: Proposed by McKay et al. (2023); code in the Qiskit Device Benchmarking repository.
European: No
How it works and its limitations Layer fidelity and error per layered gate (EPLG)
The fidelity with which a processor samples from random circuits, estimated from the ideal probabilities of the bit strings it outputs.
Maintained by: Introduced by Boixo et al. (2018); used mainly by Google Quantum AI.
European: No
How it works and its limitations Cross-entropy benchmarking (XEB)
The largest Greenberger-Horne-Zeilinger state (all qubits jointly 0 or jointly 1) a processor prepares and verifies: a test of multi-qubit entanglement.
Maintained by: No single maintainer; used by the KPI Comp 2b of the Quantum Flagship.
European: Not determined: Used as a European KPI; the benchmark itself has no single owner.
The largest random N-qubit Clifford unitary a computer runs reliably, as defined for KPI Comp 2a of the Quantum Flagship.
Maintained by: No single maintainer; used by the KPI Comp 2a of the Quantum Flagship.
European: Not determined: Used as a European KPI; the benchmark itself has no single owner.
How it works and its limitations Random Clifford unitary size
How the success of a computer falls with program width and depth, comparing structured programs with random (disordered) ones.
Maintained by: Proposed by Proctor et al. (2022).
European: No
How it works and its limitations Capability benchmarks with structured and disordered programs
Result quality and time to solution on small algorithms and applications as the problem size grows, shown on a volumetric plot of width and depth.
Maintained by: Members of the QED-C Technical Advisory Committee on Standards and Performance Metrics (repository, Apache-2.0).
European: No
How it works and its limitations QED-C application-oriented benchmarks
Application-level performance, with feature vectors that quantify how representative the chosen applications are.
Maintained by: Introduced by Tomesh et al. (2022); its code sits in the client-superstaq repository of Infleqtion (supermarq-benchmarks folder).
European: No
A shared set of benchmark circuits for evaluating compilers and other quantum software tools.
Maintained by: Technical University of Munich (Munich Quantum Toolkit), as verified for the software directory of the portal.
European: Yes: Its owner organisation is in Germany (software directory).
Performance of complete industrial applications across quantum and classical solvers, devices and simulators.
Maintained by: Open-source framework introduced by Finžgar et al. (2022) with authors from industry; maintained in its GitHub organisation.
European: Not determined: European origin not verified on the pages read (affiliations were not shown).
The largest MaxCut problem a quantum processor solves effectively with QAOA, as a hardware-agnostic and scalable score.
Maintained by: Proposed by Martiel, Ayral and Allouche (2021) as the Atos Q-score.
European: Yes: Named the Atos Q-score by its authors; EVIDEN, an Atos business, is a BACQ partner (BACQ paper).
Performance criteria for quantum computers that are meaningful for industrial users.
Maintained by: THALES, EVIDEN, CEA, CNRS, TERATEC and LNE (BACQ project).
European: Yes: A French national project (MetriQs-France).
How it works and its limitations BACQ application-oriented benchmarks
A single number N: the largest box of width N qubits and depth N^2 two-qubit gates in which every circuit of a set of QED-C-derived benchmarks succeeds.
Maintained by: IonQ (company page, version 1.0).
European: No
How well a quantum computer solves linear algebra tasks, such as linear systems, on random block-encoded matrices, as an analogue of the LINPACK benchmark of supercomputers.
Maintained by: Proposed by Dong and Lin (2021).
European: No: Both authors are at the University of California, Berkeley and Lawrence Berkeley National Laboratory (USA) (arXiv paper, first page, read on 27 October 2026).
Only initiatives verified on their own pages are listed; where a benchmarking work item could not be verified, the entry says so.
Asks for a benchmarking framework and a hardware-agnostic suite for quantum and hybrid HPC-quantum systems, with fair and reproducible comparisons, a metrics catalogue, standard formats, benchmarks from gate to application level including throughput, latency and energy, and outputs under the EUPL-1.2 licence. The EuroHPC JU page gives the topic as HORIZON-JU-EUROHPC-2024-BENCHMARK-05-02, CORDIS as HORIZON-EUROHPC-JU-2024-BENCHMARK-05-03.
Status: Closed (deadline 14 April 2026); no selected project listed on the EuroHPC JU or CORDIS pages read on 27 October 2026.
Sources: EuroHPC Joint Undertaking 2026, Call page; European Commission 2026, Programme topic page
The European standardisation committee on quantum technologies has a working group for quantum computing and simulation, the natural home of European benchmarking standards. Its entry in the standards directory.
Status: Active committee; a benchmarking work item was not verified for this release.
Sources: CEN and CENELEC 2026, Committee page (read for the shared taxonomy and the standards directory)
The international committee for quantum technology standards, including quantum computing and simulation. Its entry in the standards directory.
Status: Active committee; a benchmarking work item was not verified for this release.
Sources: IEC and ISO 2026, Committee page (read for the standards directory)
An IEEE working group whose scope is performance metrics for quantum computing hardware and software; the IEEE standards page for the project could not be read (page unavailable on 23 October 2026).
Status: Working group open to individual members; the status of the standard is not stated on its page.
Sources: IEEE Standards Association working group 2026, Working group page
A French national programme on measurements, standards and evaluation of quantum technologies, part of the national quantum strategy, which supports the BACQ application-oriented benchmarks.
Status: Described in a preprint of March 2024.
Sources: F. Barbaresco et al. 2024, arXiv:2403.12205 (March 2024)
Proposes general guidelines for quantum benchmarking that could lead to an organisation similar to the Standard Performance Evaluation Corporation (SPEC).
Status: Peer-reviewed proposal (Quantum Science and Technology, 2026).
Sources: A. Acuaviva et al. 2026, Quantum Sci. Technol. 11, 025004 (2026); arXiv:2407.10941
Which benchmarks and test algorithms the EuroHPC Joint Undertaking and the hosting sites asked for when buying quantum computers. Only what the tender documents state is shown, with the document and the page where it appears; where documents were not public or could not be read, the entry says so instead of guessing.
9 procurements, 7 with public tender documents read in full; 45 benchmarks or tests named.
Tender requirements describe what was asked for, not what the delivered machines achieved. Offers and acceptance protocols are not public.
| System | Hosting site | Platform | Benchmarks named |
|---|---|---|---|
| EuroQCS-Poland (now Piast-Q) | PSNC, Poznań (Poland) | Trapped ions (AQT, as stated in the contract announcement) | Listed on the EU tender portal but not retrievable |
| Euro-Q-Exa | LRZ (Germany) | Superconducting qubits, two systems (Part 2, Req_12) | 6 benchmarks or tests |
| EuroQCS-France (now Lucy) | GENCI and CEA, TGCC, Bruyères-le-Châtel (France) | Photonic digital quantum computer (Part 2, Req_5) | 6 benchmarks or tests |
| LUMI-Q (now VLQ) | IT4Innovations, Ostrava (Czechia) | Superconducting qubits, at least 20 in a star-shaped topology (Part 2, section 1.2 and Req_12) | 10 benchmarks or tests |
| EuroQCS-Spain | BSC, Barcelona (Spain) | Adiabatic quantum computer or annealer on superconducting qubits (Part 2, Req_3) | 5 benchmarks or tests |
| EuroQCS-Italy (now SOL) | CINECA, Bologna (Italy) | Neutral atoms, hybrid analogue and digital or fully digital; an analogue offer must include an upgrade within 18 months of acceptance, and each system must pass the Annex 8 benchmarks (Part 2, Req_4 and section 2.1) | 7 benchmarks or tests |
| EuroSSQ-HPC | SURF (Netherlands) | Semiconductor spin qubits, at least 16 (Part 2, Req_10 and Req_12) | 5 benchmarks or tests |
| MeluXina-Q | LuxProvide (Luxembourg) | Semiconductor spin qubits, at least 10 at first (Part 2, Req_17 and Req_18) | 6 benchmarks or tests |
| HPCQS quantum simulators (now Jade and Ruby) | Jülich Supercomputing Centre (Germany) and GENCI at TGCC (France) | Neutral-atom analogue simulators (Pasqal), each controlling at least 100 qubits | No tender specification with benchmarks published |
The EU Funding and Tenders portal lists the tender documents, including an Annex 8 on application performance benchmarks, but they were published on the former eTendering platform and could not be downloaded on 23 October 2026. The benchmarks are therefore not shown.
The contract announcement names no benchmark; it says that new hybrid quantum-classical use cases and benchmarks will be supported. (Advancing European Quantum Computing: Signature of the Procurement Contract for EuroQCS-Poland, Paragraph on the capabilities of the system)
Annex 8 states that thresholds are generally not set, with exceptions in the benchmark descriptions; the five required benchmarks have equal weight.
| Benchmark or test, as named | How it is used | Threshold or condition stated | On this portal | Document and place |
|---|---|---|---|---|
| Quantum volume | Required in the benchmark suite (scored under the award criteria) | None stated | Quantum volume | Euro-Q-Exa: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF pages 3 to 5, section C and benchmark 1; method: A. W. Cross et al. 2019, Phys. Rev. A 100, 032328 (2019); arXiv:1811.12926 |
| Circuit layer operations per second (CLOPS) | Required in the benchmark suite (scored under the award criteria) | None stated | CLOPS (circuit layer operations per second) | Euro-Q-Exa: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF page 5, benchmark 2; method: A. Wack et al. 2021, arXiv:2110.14108 (preprint, October 2021) |
| Creation of a GHZ state | Required in the benchmark suite (scored under the award criteria) | GHZ fidelity of at least 0.5 (genuine multipartite entanglement), on the full system or the largest qubit count that reaches it | GHZ state size | Euro-Q-Exa: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF pages 6 and 7, benchmark 3 |
| Variational quantum eigensolver (VQE) on weighted Max-Cut | Required in the benchmark suite (scored under the award criteria) | Small instances compared with classical simulation; larger ones compared with QAOA | Variational quantum eigensolver (VQE) | Euro-Q-Exa: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF pages 7 and 8, benchmarks 4 and 5 |
| Quantum approximate optimisation (QAOA) on weighted Max-Cut | Required in the benchmark suite (scored under the award criteria) | Small instances compared with classical simulation | Quantum approximate optimisation algorithm (QAOA) | Euro-Q-Exa: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF pages 7 and 8, benchmarks 4 and 5 |
| Any further test from MQT Bench | Optional, at the choice of the tenderer | None stated | MQT Bench | Euro-Q-Exa: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF page 9, section 3; method: N. Quetschlich et al. 2023, Quantum 7, 1062 (2023); arXiv:2204.13719 |
Annex 8 lists these as the mandatory minimum content of the benchmark suite; the supplier may commit further benchmarks.
| Benchmark or test, as named | How it is used | Threshold or condition stated | On this portal | Document and place |
|---|---|---|---|---|
| Fidelities of the listed single- and multi-qubit gates on every qubit and connection | Required in the benchmark suite (scored under the award criteria) | More than 1,000 runs without error mitigation; average fidelities disclosed | No page on this portal | EuroQCS-France: Annex 8, Application Performance Benchmarks, PDF page 4, section 3.1 |
| Entanglement and GHZ states | Required in the benchmark suite (scored under the award criteria) | GHZ fidelity above 0.5, on the full system or the largest qubit count that reaches it | GHZ state size | EuroQCS-France: Annex 8, Application Performance Benchmarks, PDF page 5, section 3.2 |
| Quality of the photon source, optical circuit and photodetectors, including photon loss | Required in the benchmark suite (scored under the award criteria) | Supplier proposes the tests; boson sampling runs are suggested | Boson sampling and Gaussian boson sampling | EuroQCS-France: Annex 8, Application Performance Benchmarks, PDF page 6, section 3.3 |
| Quantum LINPACK | Required in the benchmark suite (scored under the award criteria) | None stated | Quantum LINPACK, Linear systems of equations (HHL and successors) | EuroQCS-France: Annex 8, Application Performance Benchmarks, PDF pages 6 and 7, section 3.4; method: Y. Dong et al. 2021, Phys. Rev. A 103, 062412 (2021); arXiv:2006.04010 |
| Q-score (QAOA on Max-Cut) | Required in the benchmark suite (scored under the award criteria) | None stated | Q-score, Quantum approximate optimisation algorithm (QAOA) | EuroQCS-France: Annex 8, Application Performance Benchmarks, PDF page 7, section 3.5; method: S. Martiel et al. 2021, IEEE Transactions on Quantum Engineering, DOI 10.1109/TQE.2021.3090207; arXiv:2102.12973 |
| VQE energy of molecular hydrogen or lithium hydride | Required in the benchmark suite (scored under the award criteria) | Fidelity of the estimated energy levels | Variational quantum eigensolver (VQE) | EuroQCS-France: Annex 8, Application Performance Benchmarks, PDF pages 7 and 8, section 3.6 |
Annex 8 states that thresholds are generally not set; results are compared with other tenders and market studies. Measured physical characteristics must match the values committed in the offer and the minimum values of the technical specifications.
| Benchmark or test, as named | How it is used | Threshold or condition stated | On this portal | Document and place |
|---|---|---|---|---|
| Physical characteristics: qubit number, gate fidelity, connectivity, T1, T2*, qubit frequencies, readout fidelity | Required in the benchmark suite (scored under the award criteria) | Values committed in the offer, within the minimums of Part 2, section 2 | No page on this portal | LUMI-Q: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF page 3, section 4.1 |
| Quantum volume | Required in the benchmark suite (scored under the award criteria) | Must be executed and return results; committed values, if any, must be met | Quantum volume | LUMI-Q: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF pages 3 and 4, sections 4.2 and 4.2.1; method: A. W. Cross et al. 2019, Phys. Rev. A 100, 032328 (2019); arXiv:1811.12926 |
| Circuit layer operations per second (CLOPS) | Required in the benchmark suite (scored under the award criteria) | Must be executed and return results | CLOPS (circuit layer operations per second) | LUMI-Q: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF page 5, section 4.2.2; method: A. Wack et al. 2021, arXiv:2110.14108 (preprint, October 2021) |
| Volumetric testing: fidelity by circuit width and depth | Required in the benchmark suite (scored under the award criteria) | None stated | No page on this portal | LUMI-Q: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF page 6, section 4.3 |
| Oracle-based algorithm (Deutsch-Jozsa) | Required in the benchmark suite (scored under the award criteria) | None stated | No page on this portal | LUMI-Q: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF page 6, section 4.3 |
| Phase estimation | Required in the benchmark suite (scored under the award criteria) | None stated | Quantum phase estimation | LUMI-Q: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF page 6, section 4.3 |
| Hamiltonian simulation of a one-dimensional Heisenberg chain | Required in the benchmark suite (scored under the award criteria) | None stated | Hamiltonian simulation: product formulas, qubitization and QSVT | LUMI-Q: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF page 6, section 4.3 |
| Monte Carlo sampling (Woerner and Egger approach) | Required in the benchmark suite (scored under the award criteria) | None stated | Amplitude estimation and quantum Monte Carlo integration | LUMI-Q: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF page 6, section 4.3; method: S. Woerner et al. 2019, npj Quantum Information 5, 15 (2019); arXiv:1806.06893 |
| Variational algorithms (VQE and QAOA) on weighted Max-Cut, with Qiskit and MQT Bench code | Required in the benchmark suite (scored under the award criteria) | None stated | MQT Bench, Variational quantum eigensolver (VQE), Quantum approximate optimisation algorithm (QAOA) | LUMI-Q: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF pages 6 to 8, section 4.3.1; method: N. Quetschlich et al. 2023, Quantum 7, 1062 (2023); arXiv:2204.13719 |
| QED-C application-oriented benchmark software as reference implementation | Named as a reference implementation | None stated | QED-C application-oriented benchmarks | LUMI-Q: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF page 6, section 4.3; method: Members of the QED-C Technical Advisory Committee on Standards and Performance Metrics 2026, GitHub repository README (Apache-2.0) |
Annex 8 states that thresholds are generally not set; results are compared with other tenders and market studies. Every qubit must meet the T1 minimum of the technical specifications.
| Benchmark or test, as named | How it is used | Threshold or condition stated | On this portal | Document and place |
|---|---|---|---|---|
| Coherence time T1 of every qubit, with scripts for the hosting entity | Required in the benchmark suite (scored under the award criteria) | Each qubit must meet the T1 minimum of Part 2, section 2 | No page on this portal | EuroQCS-Spain: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF page 3, section 1.1 |
| Qubit frequencies, with a measurement script | Required in the benchmark suite (scored under the award criteria) | None stated | No page on this portal | EuroQCS-Spain: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF page 3, section 1.1 |
| Annealing to a good approximation of the ground state of a one-dimensional transverse-field Ising model | Required in the benchmark suite (scored under the award criteria) | None stated | Quantum annealing and adiabatic optimisation | EuroQCS-Spain: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF page 3, section 1.2 |
| Annealing schedule between two Ising or Heisenberg spin Hamiltonians, with two-qubit interaction terms on the last chip generation | Required in the benchmark suite (scored under the award criteria) | Required for the last chip generation; non-stoquastic interactions valued positively | Quantum annealing and adiabatic optimisation | EuroQCS-Spain: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF pages 3 and 4, section 1.2 |
| An application Hamiltonian with a physical or industrial justification | Optional, at the choice of the tenderer | None stated | Quantum annealing and adiabatic optimisation | EuroQCS-Spain: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF page 4, section 1.3 |
Annex 8 states that no minimum values are required unless a benchmark description says so; the set depends on whether an analogue or digital system is delivered.
| Benchmark or test, as named | How it is used | Threshold or condition stated | On this portal | Document and place |
|---|---|---|---|---|
| Sequences of random pulses on a unit-disk graph register (RSEQ), compared with emulators | Required in the benchmark suite (scored under the award criteria) | At least 5 sequences with different topologies | Analogue quantum simulation of spin and Hubbard models | EuroQCS-Italy: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF pages 6 and 7, section 1.1.2 (analogue) |
| Preparation of an antiferromagnetic state (AFMAG) on chains and square and triangular lattices | Required in the benchmark suite (scored under the award criteria) | None stated | Analogue quantum simulation of spin and Hubbard models | EuroQCS-Italy: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF pages 7 to 9, section 1.2.3 (analogue) |
| Quantum adiabatic algorithm (QAA) for maximum independent set | Required in the benchmark suite (scored under the award criteria) | At least 5 problems with different topologies and atom numbers | Analogue optimisation on Rydberg atom arrays (maximum independent set) | EuroQCS-Italy: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF pages 10 and 11, section 1.2.5 (analogue); method: S. Ebadi et al. 2022, Science 376, 1209 (2022); arXiv:2202.09372 |
| QAOA on weighted Max-Cut, compared with classical emulation | Required in the benchmark suite (scored under the award criteria) | At least 5 graph instances | Quantum approximate optimisation algorithm (QAOA) | EuroQCS-Italy: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF pages 11 and 12, section 1.2.6 (all systems) |
| Quantum volume | Required in the benchmark suite (scored under the award criteria) | None stated | Quantum volume | EuroQCS-Italy: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF page 5, section 1.1.1 (digital); method: A. W. Cross et al. 2019, Phys. Rev. A 100, 032328 (2019); arXiv:1811.12926 |
| Circuit layer operations per second (CLOPS) | Required in the benchmark suite (scored under the award criteria) | None stated | CLOPS (circuit layer operations per second) | EuroQCS-Italy: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF page 6, section 1.1.2 (digital); method: A. Wack et al. 2021, arXiv:2110.14108 (preprint, October 2021) |
| Creation of a GHZ state | Required in the benchmark suite (scored under the award criteria) | GHZ fidelity above 0.5, on the full QPU or the largest subset that reaches it | GHZ state size | EuroQCS-Italy: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF pages 9 and 10, section 1.2.4 (digital) |
Annex 8 states that thresholds are generally not defined for the technical features, but committed values must meet section 2 of the technical specifications.
| Benchmark or test, as named | How it is used | Threshold or condition stated | On this portal | Document and place |
|---|---|---|---|---|
| Number of addressable qubits; single- and two-qubit gate fidelity and speed (for example randomised benchmarking); readout fidelity and speed; T2 | Required in the benchmark suite (scored under the award criteria) | Committed values, within the minimums of Part 2, section 2 | Randomised benchmarking family | EuroSSQ-HPC: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF page 3, section 1; method: E. Magesan et al. 2011, Phys. Rev. Lett. 106, 180504 (2011); arXiv:1009.3639 |
| Creation of a GHZ state | Required in the benchmark suite (scored under the award criteria) | GHZ fidelity above 0.5, on the full QPU or the largest subset that reaches it | GHZ state size | EuroSSQ-HPC: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF pages 4 and 5, section 2.1 |
| Variational algorithms on Max-Cut, scored with the Q-score protocol (100 random graphs per size) | Required in the benchmark suite (scored under the award criteria) | Q-score test passed when beta(n) exceeds 0.2 | Q-score, Variational quantum eigensolver (VQE), Quantum approximate optimisation algorithm (QAOA) | EuroSSQ-HPC: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF pages 5 and 6, section 2.2; method: S. Martiel et al. 2021, IEEE Transactions on Quantum Engineering, DOI 10.1109/TQE.2021.3090207; arXiv:2102.12973 |
| Bernstein-Vazirani hidden bit string | Required in the benchmark suite (scored under the award criteria) | Success probability reported for each qubit count from 2 plus an ancilla | No page on this portal | EuroSSQ-HPC: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF page 6, section 2.3 |
| QED-C benchmarks, PSNC OpenQBench and MQT Bench as reference implementations | Named as a reference implementation | None stated | QED-C application-oriented benchmarks | EuroSSQ-HPC: Annex 8, Performance Benchmarks and Technology Roadmap Assessment, PDF page 4, section 2; method: Members of the QED-C Technical Advisory Committee on Standards and Performance Metrics 2026, GitHub repository README (Apache-2.0) |
Runtime is the median total time over 100 iterations of 100 shots, from transpiling to the result; fidelity is the median Hellinger-type fidelity against an exact simulation. The circuits are fixed OpenQASM 2 files generated with MQT Bench.
| Benchmark or test, as named | How it is used | Threshold or condition stated | On this portal | Document and place |
|---|---|---|---|---|
| Quantum volume | Mandatory requirement with a minimum | At least 2^5 = 32 (mandatory requirement of the initial QPU) | Quantum volume | MeluXina-Q: Tender specifications, Part 2, technical specifications, PDF page 9, section 2.1, Req_22; method: A. W. Cross et al. 2019, Phys. Rev. A 100, 032328 (2019); arXiv:1811.12926 |
| GHZ state, 3 and 10 qubits | Required in the benchmark suite (scored under the award criteria) | Runtime and fidelity reported; committed results must be achieved at acceptance | GHZ state size | MeluXina-Q: Annex 8, Application Performance Benchmarks, PDF pages 3 to 5, section C, BM1 |
| Quantum Fourier transform, 4 and 10 qubits | Required in the benchmark suite (scored under the award criteria) | Runtime and fidelity reported; committed results must be achieved at acceptance | MQT Bench | MeluXina-Q: Annex 8, Application Performance Benchmarks, PDF pages 3 to 5, section C, BM2 |
| Quantum phase estimation, 4 and 10 qubits | Required in the benchmark suite (scored under the award criteria) | Runtime and fidelity reported; committed results must be achieved at acceptance | MQT Bench, Quantum phase estimation | MeluXina-Q: Annex 8, Application Performance Benchmarks, PDF pages 3 to 5, section C, BM3 |
| Grover search, 4 and 6 qubits | Required in the benchmark suite (scored under the award criteria) | Runtime and fidelity reported; committed results must be achieved at acceptance | MQT Bench, Grover search and amplitude amplification | MeluXina-Q: Annex 8, Application Performance Benchmarks, PDF pages 3 to 5, section C, BM4 |
| Benchmark circuits taken from MQT Bench; fidelity metric from the QED-C paper | Named as a reference implementation | None stated | MQT Bench | MeluXina-Q: Annex 8, Application Performance Benchmarks, PDF page 5, sections C and D; method: N. Quetschlich et al. 2023, Quantum 7, 1062 (2023); arXiv:2204.13719 |
No tender specification with benchmarks was found among public EuroHPC or EU tender documents; the announcement names only the capability to control at least 100 qubits and lists benchmarking and certification as future project work.
Selection is described as requiring simulators able to control at least 100 qubits; benchmarking and certification or performance analysis are named as later HPCQS work, not as selection criteria. (With two 100 qubits quantum computers from PASQAL, FZJ and GENCI boost HPCQS, the pan-European hybrid HPC/quantum infrastructure, Opening paragraphs and the description of the HPCQS work)
Olivier Ezratty, Understanding Quantum Technologies (CC BY-NC-ND 4.0). Links only; no text of the book is reproduced on this portal.