Run your own quantum hardware benchmarks
Every result below states its circuit, environment and method, including the one we withdrew. The scripts are open source, so you can run the same workloads on your own hardware and accounts and compare the numbers yourself.
Results with a recorded method
We list a result here only when it comes with the circuit, the environment and the method needed to reproduce it. The one optimiser result we published was withdrawn; it stays listed so you can see why. Click a metric for the details.
Autopilot sample efficiency: withdrawn
QCOS Autopilot versus SPSA on a GHZ-6 circuit (CPU state-vector simulator), withdrawn on 25 August 2026
Not published yet
The following are not listed because no recorded run backs them today. We make no performance, fidelity or availability claim for them.
- Hardware fidelity and quantum volume tables
- Latency percentiles and ledger overhead
- Budget-enforcement accuracy
- Backend comparison and price tables
- Circuit-depth reduction across benchmark suites
What we can state today: QCOS state-vector simulation on CPU runs up to 24 qubits, and each run records an evidence bundle.
Benchmark Methodology
Reproducibility
Runs executed through QCOS record an evidence bundle sealed under a SHA-256 Merkle root, verifiable offline with softquantus-verify.
Transparency
Each listed metric states its circuit, environment, baseline and sample size. Click a metric to view them.
Scope
The only result listed is measured on a CPU state-vector simulator. It says nothing about any quantum processor.
Run the benchmarks yourself
All benchmark scripts are open source. Run them on your own hardware and judge the results on your own workload.