QONTOS Benchmarks

Reproducible evidence, methodology, and regression validation for the QONTOS platform.

The public proof layer for correctness claims across the SDK, simulator, and modular execution stack.

Overview · Installation · Quick Start · Docs Hub · Benchmark Suite · Methodology · Report Format

QONTOS Benchmarks is the public evidence repo for the QONTOS platform. It contains benchmark definitions, execution methodology, report schemas, and reproducible outputs used to validate correctness, performance, and regression behavior across the public QONTOS stack. This repository is the reference point for public technical claims made by QONTOS.

Start with docs/index.md for the lightweight docs hub and the key benchmark evidence pages. For the canonical install and release policy across the public repos, use the shared policy.

pip install "qontos-bench @ git+https://github.com/qontos/[email protected]"

Note: Once published to PyPI, this will simplify to pip install qontos-bench.

git clone https://github.com/qontos/qontos-benchmarks.git
cd qontos-benchmarks
pip install -e ".[dev]"
pytest

CI validates both paths: the released-tag install (on main push) and the local-checkout install (on every PR). CI also now emits hybrid, hybrid-stress, and transduction-closure benchmark artifacts from the live benchmark packs so downstream systems workflows can consume real generated JSON reports instead of static fixtures. The repo now also ships a separate hybrid stress pack and a dedicated transduction closure pack so we can probe the superconducting-photonic seam with workloads that specifically target transduction, retry, calibration, phase stability, logical-patch transport, control-jitter, and entanglement-supply pressure.

# Run the full benchmark suite
python -m qontos_bench

# Run a specific benchmark
python -m qontos_bench --benchmark bell_pair

# Generate a report
python -m qontos_bench --report json --output results.json

• photonic-bell
• teleport
• remote-cnot
• distributed-ghz
• syndrome-burst

• entanglement-swap
• teleport-ladder
• remote-parity
• distributed-ghz-ladder
• patch-syndrome
• transducer-cal
• logical-patch-handoff

• transduction-closure
• transducer-cal
• logical-patch-handoff

Run the stress pack directly:

python -m qontos_bench --circuit hybrid-stress

Run the closure pack directly:

python -m qontos_bench --circuit transduction-closure

Each benchmark is executed through the full QONTOS pipeline:

1. Circuit normalization via CircuitNormalizer
2. Execution via LocalSimulatorExecutor (Qiskit Aer)
3. Fidelity computation: F = (expected state counts) / total_shots

Pass threshold: 0.85. Default shot count: 8,192.

The report schema now also includes a stressor_summary section derived from benchmark metadata so downstream systems workflows can see which hybrid bottleneck classes are still weak, not just which circuit families passed.

These benchmarks validate pipeline correctness — that the ingest-normalize-execute path preserves circuit semantics. They are designed to detect regressions, not to measure hardware noise characteristics.

• Backend: Qiskit Aer AerSimulator (noiseless statevector)
• Shot count: 8,192 (configurable)
• Fidelity definition: fraction of shots landing in expected states
• All results are simulator-based until hardware QPU integration

{
  "timestamp": "2026-03-23T12:00:00Z",
  "backend": "aer_simulator",
  "shots": 8192,
  "benchmarks": [
    {
      "name": "bell_pair",
      "qubits": 2,
      "fidelity": 1.0,
      "passed": true,
      "counts": {"00": 4096, "11": 4096}
    }
  ]
}

Apache License 2.0

Built by Zhyra Quantum Research Institute (ZQRI) — Abu Dhabi, UAE