Home » Quantum Computing » Qunova’s HI-VQE Hits Chemical Accuracy at 80 Qubits

Qunova’s HI-VQE Hits Chemical Accuracy at 80 Qubits

A South Korean startup says its quantum algorithm matched a classical benchmark on a biologically crucial iron-sulfur cluster, using far less classical computing than its earlier attempts. Independent labs haven’t…

Abstract glowing cube cluster symbolizing Qunova's chemical accuracy result on an iron-sulfur cluster

In October 2024, Qunova Computing needed 20 qubits to reach chemical accuracy with its chemistry software. Two years later, the South Korean company needed four times that many qubits for its hardest molecule yet. In exchange, it says, classical supercomputers needed roughly 300 times less help to check the answer. Qunova reports that its HI-VQE algorithm matched a trusted classical benchmark on an 80-qubit model of an iron-sulfur cluster. That molecule sits at the core of enzymes every living thing depends on. The claim comes from the company’s own announcement and a technical paper posted to arXiv. It has not yet cleared peer review.

What Chemical Accuracy Actually Means

Chemists have a specific bar for a useful prediction. They call it chemical accuracy. It means landing within 1.6 milliHartree of the true answer, a unit so small it only matters at the scale of individual molecules. Miss that bar, and a simulation can get a reaction’s speed or a material’s stability badly wrong. Hit it, and the simulation becomes something a chemist can actually plan around.

Qunova says its extrapolated result for the molecule landed about 1.1 thousandths of a Hartree from the classical reference value. That’s comfortably inside the 1.6 milliHartree bar. The reference itself came from a method called DMRG, density matrix renormalization group. It’s one of the more trusted tools chemists already have for checking tricky cases like this one. Other teams are chasing the same bar from a different angle: Qedma’s error-mitigation software recently pushed a simple water molecule inside that threshold on existing IBM hardware, without changing the underlying algorithm at all.

Four Years, Four Escalating Claims

This is not Qunova’s first chemical accuracy announcement, and each one has quietly raised the stakes.

When Hardware What Was Tested Claim
October 2024 20-qubit IQM, 24-qubit IBM Eagle Lithium sulfide First chemical accuracy on commercial quantum hardware
January 2026 Up to 56 qubits (various) General validation for AWS listing Ready for commercial Braket users
November 2026 80-qubit IBM Quantum System Two + RIKEN’s Fugaku [4Fe-4S] iron-sulfur cluster Chemical accuracy with roughly 300x less classical compute

Sources: Qunova Computing press releases, Research & Development World, arXiv:2601.06935.

The jump from small test molecules to an iron-sulfur cluster is the part worth noticing. Qunova is one of several teams chasing the same molecule right now. A separate Caltech, IBM and RIKEN team published its own attempt at a similar cluster in Science Advances in 2025. That run used 77 qubits and landed an energy estimate that beat older classical shortcuts, but it still fell short of matching DMRG. Qunova’s newer claim, if it holds up, would close that remaining gap. Pairing a quantum chip with a classical supercomputer for chemistry is becoming its own trend: Japan’s University of Osaka recently wired a quantum chip directly into its OCTOPUS supercomputer for the same kind of hybrid workflow, though on a far smaller test molecule.

Why a Cluster of Iron and Sulfur Is Such a Hard Test

The molecule in question, called [4Fe-4S], is a small cube of four iron atoms and four sulfur atoms. Versions of it sit inside nitrogenase, the enzyme bacteria use to pull nitrogen out of the air and turn it into a form plants can use. Related clusters also appear in hydrogenase enzymes, which some microbes use to produce or consume hydrogen gas.

Both processes matter well beyond a lab bench. The industrial version of nitrogen fixation, the Haber-Bosch process, already makes the fertilizer that feeds a large share of the planet. But it runs at high heat and pressure, and it burns through enormous amounts of energy. Bacteria do the same job at room temperature using a cluster like this one. Chemists have wanted an accurate quantum-mechanical picture of that trick for decades. Part of the hope is that it could point toward cheaper, lower-energy ways to make fertilizer or produce hydrogen fuel. Electrons in iron-sulfur clusters interact in ways that confuse standard approximations. That’s exactly why researchers keep using this molecule as a stress test for new methods, quantum or classical.

Why It Matters to You

Nobody gets cheaper fertilizer or cleaner hydrogen from this announcement. It is a benchmark result on a small model system, run by one company, not yet checked by independent reviewers. The realistic path from here to your grocery bill or your utility rates runs through years of further research connecting results like this one to an actual industrial catalyst. What this result does show is a specific, falsifiable claim: a defined molecule, a defined hardware pairing, and a number anyone with the right training can try to reproduce. That is a more useful kind of progress than a vague promise, even if the payoff is still a long way off.

What We Don’t Know Yet

  1. The 1.6 milliHartree chemical accuracy threshold comes from Qunova’s own framing. Other teams in the field use similar thresholds, but there is no universal regulator that certifies the term.
  2. Both the press release and the arXiv paper come from Qunova and its academic co-authors, including RIKEN scientists. No outside lab has independently reproduced the 80-qubit result yet.
  3. The paper itself has been revised once already, in September 2026, to add results beyond its original January 2026 version. A preprint that keeps growing is normal in this field, but it is still not the same as a published, peer-reviewed paper.
  4. The roughly 300x cost reduction is measured against Qunova’s own prior quantum-centric supercomputing runs, not against every competing method in the field.
  5. This is a model cluster studied on its own, not the full nitrogenase or hydrogenase enzyme in a real biological setting, which is vastly more complex.

Qunova says it next wants to add orbital optimization to HI-VQE, aiming at molecules with far more possible electron arrangements than this one. Whether a bigger, less noisy quantum chip gets there before classical methods like DMRG simply get faster is the more interesting race hiding behind this specific result.

Sources

Related Reading