Circuit Layer Operations per Second
Circuit Layer Operations per Second (CLOPS) is a benchmarking metric of quantum computers developed by IBM in 2021.[1] While Quantum volume is more indicative of a qubit's performance and quality, CLOPS is a measure of a quantum computer’s execution speed. CLOPS incorporates both the time to run a circuit as well as the real-time and near-time classical computation, enabling it to serve as a holistic single measure of speed. CLOPS in conjunction with the number of qubits and the quantum volume was originally envisioned by IBM to be the 3 core metrics for comparing quantum devices.[2]
Redefinition
[edit]Originally CLOPS was designed to be related to quantum volume, in the sense that each circuit layer was idealized square quantum volume circuits.[3] However, in 2023 IBM created a new definition of CLOPS which is related to a specific hardware layer. The different versions are differentiated with a subscript: CLOPSv for virtual and CLOPSh for hardware. Functionally, the core difference is that instead of using quantum volume in the calculation, a different metric known as layer fidelity is used instead, measured by the error per layered gate (ELPG) protocol. This enables different hardware with different quantum volumes to be more accurately compared.[4]
Mathematical explanation of the benchmark
[edit]- Define M template circuits; each one composed of K different quantum circuits
- S is the number of shots to run
- D is the quantum volume as log2
- T is the total amount of time in seconds taken for execution, including the data transfer, transpilation, measurements, resets, and delays
CLOPSv = [5]
The main update in CLOPSh was to replace the quantum volume metric D above with the something IBM termed the layer fidelity, which according to IBM "provides a benchmark that encapsulates the entire processor’s ability to run circuits while revealing information about individual qubits, gates, and crosstalk." Layer fidelity is measured by the error per layered gate (EPLG). [4] This inflates the CLOPS score for quantum computers that are sparsely connected.
Table of measured CLOPS
[edit]| Date | Company | QPU | CLOPSv | CLOPSh |
|---|---|---|---|---|
| October 2021[6] | IBM | Bogota | 1419 | |
| October 2021[6] | IBM | Toronto | 951 | |
| October 2021[6] | IBM | Brooklyn | 753 | |
| August 2023[7] | Qutech | Starmon-5 | 372 | |
| August 2024[8] | IQM Quantum Computers | Garnet | 2600 | |
| February 2025[6] | IBM | Heron r2 | 2400 | 195000 |
| February 2025[6] | IBM | Heron r1 | 2400 | 210000 |
References
[edit]- ^ Gandharv, kumar (2021-11-03). "IBM Proposes New Metric To Measure Quantum Computer Performance | AIM". Analytics India Magazine. Retrieved 2025-10-18.
- ^ "Driving quantum performance | IBM Quantum Computing Blog". www.ibm.com. Retrieved 2025-10-18.
- ^ AbuGhanem, Muhammad (19 August 2025). "Superconducting quantum computers: who is leading the future?". EPJ Quantum Technology. 12 (1) 102: 1–75. Bibcode:2025EPJQT..12..102A. doi:10.1140/epjqt/s40507-025-00405-7. ISSN 2196-0763.
- ^ a b "Updating how we measure quantum quality and speed | IBM Quantum Computing Blog". www.ibm.com. Retrieved 2025-10-18.
- ^ "Circuit Layer Operations Per Second (CLOPS) benchmark".
- ^ a b c d e "Compute resources | IBM Quantum Platform". quantum.cloud.ibm.com. Retrieved 2025-10-18.
- ^ Schoot, Ward van der; Wezeman, Robert; Eendebak, Pieter Thijs; Neumann, Niels M. P.; Phillipson, Frank (2023-10-02), "Evaluating three levels of quantum metrics on quantum-inspire hardware", Quantum Information Processing, 22 (12) 451, arXiv:2310.01120, Bibcode:2023QuIP...22..451V, doi:10.1007/s11128-023-04184-x, retrieved 2025-10-18
- ^ Abdurakhimov, Leonid; Adam, Janos; Ahmad, Hasnain; Ahonen, Olli; Algaba, Manuel; Alonso, Guillermo; Bergholm, Ville; Beriwal, Rohit; Beuerle, Matthias (2024-08-22), Technology and Performance Benchmarks of IQM's 20-Qubit Quantum Computer, arXiv:2408.12433, retrieved 2025-10-18