When a quantum circuit returns wrong results, the question is always the same: was it the hardware, or the code? Q-Verdict answers it.
The free client is in early access. Request a spot and we'll provide installation instructions and an access key.
Annual licensing for qualifying organizations. We review each request. Free client docs: doc.html →
Current error mitigation techniques (zero noise extrapolation, probabilistic error cancellation) are source-agnostic. They attempt to invert the noise channel regardless of whether the error comes from hardware or from a bug in the circuit. If a software bug is present, these methods will extrapolate the wrong result to the zero-noise limit.
Developers have no tool to answer the prior question: is this a bad result from correct code running on noisy hardware, or a bad result from a circuit that has a logical error? Until that question is answered, error mitigation is answering the wrong question. Q-Verdict answers it.
The circuit logic is correct. The bad result is consistent with hardware-level error patterns: readout calibration drift, coherent gate error, cross-talk, or decoherence accumulation. Re-run with error mitigation, adjust calibration, or increase shots.
The circuit contains a logical error that would produce incorrect results even on a perfect device. Common causes: incorrect qubit ordering, missing gates, phase errors, or transpiler-introduced discrepancies. Fix the circuit before using QPU time.
The evidence is insufficient to distinguish hardware from code error at this shot count or noise level. Q-Verdict reports what it found and what additional data would resolve the classification.
The free client runs entirely offline. No data leaves your machine. No sign-up required.
# Install pip install q-verdict # Analyze a Qiskit circuit from qverdict import QVerdict from qiskit.circuit.library import QFT circuit = QFT(4) result = QVerdict().analyze(circuit, shots=2048) print(result.verdict) # "noise" | "bug" | "inconclusive" print(result.reason) # plain-language explanation print(result.evidence) # structured diagnostic data
Early access: PyPI release pending. Install instructions provided to approved participants.
Pass any supported circuit type directly. No circuit rewriting required.
The licensed engine is available two ways. On-prem: an encrypted, license-gated package that runs in-process on your own hardware, so circuits and results never leave your environment. Built for defense, finance, and other teams that can't send work to a third-party service. Hosted: a connected service where your circuits are submitted to the Q-Verdict API and the structural verdict is returned to your pipeline; the engine's proprietary implementation stays on our infrastructure and is never distributed. Hosted access is currently limited while we complete evaluation with a small number of teams.
| Capability | Free (Apache 2.0) | Enterprise (on-prem or hosted) |
|---|---|---|
| Noise vs. bug classification | Included | Included |
| Framework adapters (Qiskit, Cirq, TKET, Braket, OpenQASM) | Included | Included |
| Offline execution, no data egress | Included | On-prem: included · Hosted: requires API connection |
| Structural verdict (catches what a histogram misses, no reference run needed) | Not included | Included |
| Photonic and continuous-variable circuit support | Not included | Included |
| Hosted, connects directly into your pipeline (access currently limited) | Not included | Included |
| Team workspace, usage dashboard, dedicated support, custom SLA | Not included | Included |
The enterprise engine is available on-prem or hosted (access currently limited) to qualifying organizations running quantum programs at scale: hardware vendors, national labs, and large technology and financial institutions with active QPU programs.
Annual licensing. We review each request before scheduling a conversation. Include your organization, QPU platform, and use case.