Science publishes cross-platform quantum error-correction benchmark to guide public funding and procurement

A consortium coordinated by the US National Institute of Standards and Technology published today in Science the first open benchmark for error-corrected quantum computation across multiple hardware platforms. The study measures logical error rates, overhead, and latency for the same set of small fault-tolerant circuits on trapped-ion, superconducting, and neutral-atom systems in North America and Europe, aiming to give engineers and funders a common yardstick.

The group also releases a reporting format and reference datasets that US and European agencies say they will evaluate in upcoming calls and testbeds. With new public quantum budgets moving into multi-year procurement, the work arrives as program managers seek defensible milestones that are comparable across vendors.

The authors, including Kenneth R. Brown at the Georgia Institute of Technology, Barbara M. Terhal at QuTech and TU Delft, and Travis S. Humble at Oak Ridge National Laboratory, define three task classes. These are Clifford stabilizer circuits for syndrome extraction, distance-scaled logical memory, and a small error-detected arithmetic kernel. Each was executed using surface code or subsystem code layouts with explicit accounting of calibration time, idle error, and control bandwidth.

What was measured was not headline gate fidelities, but end-to-end logical behavior under a fixed protocol. The team reports that no platform achieved sustained logical break-even under continuous operation. When normalized by physical qubit quality and available parallelism, the spread in logical error per cycle across the three modalities narrowed to within one order of magnitude for the tested code distances. That suggests similar engineering bottlenecks, primarily in leakage management, crosstalk under concurrent control, and decoder latency.

Data contributions came from industrial and academic labs, including IBM Research for superconducting circuits, Quantinuum and the University of Innsbruck for trapped ions, and QuTech and Harvard affiliated teams working with neutral atoms. The paper appears in Science as a Report, with all raw data, decoders, and circuit definitions posted in an open repository maintained by NIST.

For engineers, the immediate change is a move from device-level metrics to task-level acceptance tests that can be reproduced and audited. For policymakers, the checklist and datasets provide a basis for milestone language in grants and contracts that focuses on logical performance, energy per logical operation, and throughput under load.

Standardized tests rarely settle platform debates, but they shape them. If agencies adopt these benchmarks in procurement, quantum hardware roadmaps will likely align around decoder integration, control parallelism, and thermal budgets, rather than only pushing single gate fidelities.