A shared benchmark suite for logical qubits gets a first draft
Vendors publish incomparable numbers. A working group wants one protocol everyone can run.
By Ada Renner
Written in-house by the Working group desk — an explainer, not a report of a news event.
- Protocol covers logical error rate, cycle time, and overhead together.
- Results must include the decoder and its latency.
- Reference implementation is public and runs on simulators.
Comparing quantum processors today means comparing marketing. Vendors report different metrics under different assumptions, and almost none of them include decoder latency.
A cross-institution working group has published a draft benchmark that reports logical error rate, cycle time, and physical overhead as one bundle, and requires disclosure of the decoder used.
A reference implementation runs against simulators, which lowers the barrier for smaller groups to publish comparable numbers.
This story started as a thread in our Discord. Join the discussion with the researchers and engineers who work on it.
Related
Open-source compiler cuts logical-qubit overhead by a third
A new lattice-surgery scheduler squeezes routing overhead out of fault-tolerant circuits, and it ships under a permissive license.
Surface-code processor stays below threshold as it scales to 1,000 qubits
The logical error rate keeps falling as the code distance grows — the clearest sign yet that adding qubits now buys reliability instead of noise.
Trapped-ion shuttling record pushes two-qubit fidelity past 99.9%
Moving ions between zones used to cost coherence. A revised trap geometry makes transport nearly free.
