About Q-Advantage

Independent benchmarks for the quantum era.

The quantum industry runs on press releases, roadmaps, and selectively reported milestones. Q-Advantage runs on GitHub Actions logs. We build independent measurement infrastructure for three questions: how close current hardware is to breaking today's cryptography, how the post-quantum replacements perform on real silicon, and which machines can execute which algorithms with what fidelity.

Every spec is sourced. Every benchmark re-runs on a schedule. Every result links back to the public Actions workflow that produced it.

What we measure

Three benchmarks. One discipline.

Each product targets a different layer of the quantum-cryptography problem. The discipline is the same across all three: vendor-published or peer-reviewed sources only, every figure dated and methodology-tagged, every result reproducible from a public commit.

Q-Day Index
The threat horizon
Live
What it measures

A 0–100 score for how close today's quantum hardware is to breaking RSA-2048, computed against a named, published resource estimate (Gidney 2025, currently <1M physical qubits).

How

A multiplicative-gate score: distance is governed by logical-qubit capacity, two-qubit gate fidelity at the fault-tolerance threshold, and a multiplier rewarding a demonstrated below-threshold error-correction result. Every system's inputs come from a vendor data sheet or peer-reviewed paper, with the measurement method (XEB vs ECR vs randomized benchmarking) surfaced alongside the value. The 'readiness' axis — preconditions assembled toward breaking RSA — is shown as a separate, structurally different visual so the two are never confused.

Status

Eight scored systems plus two analog (N/A) entries and four footnoted candidates. Frontier sits in the low single digits today; the trajectory is the story.

Open Q-Day Index
Q-Shield
PQC benchmarks · daily
Live
What it measures

Independent performance benchmarks for the NIST-standardized post-quantum cryptographic algorithms — ML-KEM (FIPS 203), ML-DSA (FIPS 204), and SLH-DSA (FIPS 205).

How

Each algorithm is benchmarked across keygen, encap/decap (KEMs) and sign/verify (signatures) on a self-hosted GitHub Actions runner. CPU pinned to a single core, garbage collection disabled during measurement, 1,000 timed iterations per operation after 50 warmups. Full environment captured per run — CPU model, kernel version, liboqs version, git SHA, instance type — so every result is reproducible bit-for-bit. CPU steal-time and load average are captured so burstable-instance throttling is visible in the audit trail.

Status

Re-runs daily at 06:00 UTC. Every data point on the dashboard deep-links to the exact Actions run that produced it.

Open Q-Shield
Q-Arena
Algorithm leaderboard
In preview
What it measures

Real quantum algorithms executed on real quantum hardware — the same circuit run across multiple machines so machines can be compared on what they actually do, not what their data sheets claim.

How

Standardised circuits (initially Bell-state preparation and small entanglement-witness tests) run on AWS Braket simulators and on real hardware via the IBM Quantum free tier. For each (circuit, machine) pair: success rate, gate count, queue time, and circuit depth recorded. Real-hardware results compared against simulator-ideal so noise can be isolated from algorithmic structure.

Status

In design. Exact circuit list and machine roster will be finalised as we begin running them — anything stated here in advance is a plan, not a promise.

How it works

Reproducibility is the product.

Every chart, score, and comparison on this site is the output of code in a public repo. The recipe, the run, the artifacts, and the result are linked end-to-end so anyone can audit, re-run, or fork what we publish.

01
Recipe in the repo.
The benchmark scripts, the scoring engine, and the dataset live in the public q-advantage repo. Every figure on the dashboard traces back to a versioned file.
02
Runs on real hardware.
Q-Shield benchmarks execute on a self-hosted GitHub Actions runner with CPU pinning and GC disabled during measurement. Q-Arena targets cloud quantum hardware (AWS Braket, IBM Quantum).
03
Public Actions logs.
Every benchmark run is a public GitHub Actions workflow. Logs and artifacts are retained for 90 days; periodic database snapshots will be published as public releases for longer auditability.
04
Every result links back.
Every data point on the dashboard carries a deep-link to the exact Actions run that produced it. Click a sparkline dot, hit the run page, inspect the logs. No black box.
Browse workflow runsView benchmark recipesSource repository
Frequently asked

Questions worth asking.

What is Q-Advantage?
An independent, vendor-neutral benchmarking platform for the quantum era. It publishes three measurement products — the Q-Day Index (distance to breaking RSA-2048), Q-Shield (PQC algorithm performance), and Q-Arena (real algorithms on real quantum hardware). Every result is sourced, reproducible, and auditable end-to-end on GitHub.
Who runs Q-Advantage?
Q-Advantage is currently built without external funding. The codebase, data, methodology, and Actions logs are open for inspection. There is no paid tier. If that ever changes — sponsorship, subscriptions, customer engagements — it will be announced publicly with the terms in writing.
What does “vendor-neutral” mean in practice?
No quantum hardware or PQC software vendor pays for placement, ranking, or early access. The Q-Day Index hero is the field frontier, not a named winner; the table lists every machine with its own score and inputs. Every spec carries its source, measurement method, and confidence level so readers can judge the data, not trust ours.
Why does the Q-Day Index threat score sit in the single digits?
Because that is the honest reading today. The score is a multiplicative gate — it goes to zero whenever a system lacks demonstrated, standing, error-corrected logical qubits. Only one system in the dataset clears that bar at the time of writing. The trajectory is the story, not the absolute digit, and the readiness column tracks that trajectory.
Is there a projected Q-Day year?
Not yet. A projection is only as good as its model, and we have not built one that survives hostile inspection. When we publish one, the model, the assumptions, and the inputs will all be open. Anything else would be a guess dressed up as analysis.
How do you handle systems that don't fit the schema?
Analog Hamiltonian simulators (QuEra Aquila, Pasqal Orion Alpha) have no gate-model two-qubit fidelity, so they appear in the table as explicit N/A — a category difference, not a low score. Photonic and silicon-spin systems with peer-reviewed credentials but no deployed multi-qubit processor appear as footnotes with the specific reason they fall short of the scoring bar.
What is the sourcing bar for the dataset?
Vendor-published or peer-reviewed only. Every numeric field carries its URL, its confidence level (peer-reviewed vs vendor-published), the measurement method (XEB vs ECR vs randomized benchmarking, etc.), and the date the figure was true. Press releases, blog posts, and secondary aggregators do not clear the bar. Where they conflict with published values, the published value wins and the gap is noted — for example, Google's blog says “approaching 100 µs” coherence, the Nature paper says 68 µs; we cite the paper.
Can I challenge a number?
Yes — please. There is a feedback form at the bottom of the Q-Day Index page, and a public GitHub Issues tracker for anything you want a public record of. Concrete corrections improve every subsequent run.
Can I re-run a benchmark myself?
Yes. The Q-Shield benchmark script (benchmark/benchmark.py) is a standalone Python program; clone the repo, install liboqs at the matching version, and it produces the same JSON. The Q-Day Index emitter (benchmark/build-q-day-index.py) runs against the committed dataset and reproduces the dashboard JSON deterministically.
Is everything open source?
The benchmark code, the scoring engine, the dataset, the methodology document, and the frontend are all in the public q-advantage repository.
Contribute

Public critique is how this stays honest.

If you spot a wrong number, a weak source, a missing system, or a method tag we have misapplied — say so. If you want to discuss the scoring formula, the anchor target, or whether a vendor's claim clears the bar — open an issue.

GitHub organisationOpen an issuehello@qadvantage.io