QEncode is an open-source benchmark standard and certification platform for reproducible VQE quantum chemistry evaluation. Our goal is to be the accepted standard for quantum algorithm performance — the same role that MLPerf plays in classical machine learning.
Quantum algorithm results published today are almost impossible to compare across papers and platforms. Different molecule choices, basis sets, active space definitions, encodings, ansatz settings, optimizer parameters, and runtime environments mean that two papers claiming to solve the same problem are often solving completely different ones. QEncode fixes this with a fixed benchmark suite, a single reference standard (CASCI at cc-pVDZ), and signed certification artifacts that make every result independently verifiable.
47 certified entries
Across 16 molecules at cc-pVDZ basis — H₂, HF, LiH, BeH₂, H₂O, NH₃, H₂CO, C₄H₆, (H₂O)₂, C₄H₄, H₄, N₂, H₆, benzene, H₈, and H₁₀.
N₂ certified — 4.5 mHa gap
12 qubits tapered to 8, CASSCF orbital optimization, certified via the hardware-efficient ansatz with L-BFGS-B analytic gradients. DARPA QB-GSEE benchmark candidate.
Benzene certified — first aromatic
[6e,6o] π active space, D6h symmetry, CASSCF orbitals, certified at an 8.741 mHa best gap (hardware-efficient ansatz with L-BFGS-B; also certified via ADAPT-VQE at 9.540 mHa) — direct relevance to pharmaceutical chemistry.
H₁₀ certified — largest system in the suite
[10e,10o] active space, 20 JW qubits tapered to 18, CASSCF orbitals, certified via ADAPT-VQE at a 9.977 mHa best gap. DARPA QB-GSEE aligned.
Ed25519-signed artifacts
Every certified entry carries a cryptographic signature and SHA-256 provenance hash.
Fully open data
All entries, reference energies, and run configs published in the public GitHub repository.
DARPA's Quantum Benchmarking program (QB-GSEE) explicitly targets ground-state energy estimation for molecules where classical methods are insufficient. QEncode's certified N₂ result at cc-pVDZ with a [6e,6o] active space is directly comparable to the QB-GSEE target specification — with a provenance hash and Ed25519 signature that makes the claim verifiable.
As quantum hardware moves toward fault tolerance, the field needs a trusted benchmark that hardware teams, algorithm researchers, and program officers can all point to. QEncode is building that infrastructure now, starting with near-term VQE on classical simulators and extending to real hardware certification in v4.4.
Every number comes from the real pipeline: PySCF computes the reference, PennyLane runs the circuit, gap = |E_VQE − E_CASCI| computed from real outputs. If a molecule fails to converge, the result is recorded as-is — never discarded or adjusted.
The full benchmark specification, execution scripts, and requirements files are open source. Anyone can reproduce any entry with a single command. The certification layer adds signatures and managed execution — it does not change the methodology.
The self-run path is free and always will be. Managed certification — with signed artifacts, provenance receipts, and audit-ready reports — is available for teams that need verified results for publications, grant applications, or hardware evaluations.
Entries that don't pass certification (gap ≥ 0.01 Ha) appear in the Research tab, not the trash. The N₂ hardware-efficient result — 0.139 Ha with a gradient-free optimizer — sits in the database beside the 4.5 mHa result from the same ansatz with analytic gradients. And when the v4.4 re-baseline demoted N₂ UCCSD from certified to research tier, the entry stayed and the change was published, rather than quietly disappearing.
Independent, signed proof that your quantum algorithm or hardware outperforms classical methods on real chemistry benchmarks — accessible to every researcher, trusted by every reviewer.
Questions? [email protected]
