## Does Quantum Optimization Actually Beat Classical Solvers? QOBLIB Now Provides the Scorecard

A consortium anchored by [IBM Quantum](https://quantumintel.tech/companies/ibm), Zuse Institute Berlin, Technische Universität Berlin, and Purdue University has published the Quantum Optimization Benchmarking Library — QOBLIB — in *Nature Computational Science*, establishing the field's most comprehensive open standard for measuring whether quantum or [hybrid quantum-classical](https://quantumintel.tech/glossary/hybrid-quantum-classical) algorithms can outperform state-of-the-art classical solvers on combinatorial optimization. The library launches with 1,264 curated problem instances spanning ten NP-hard problem classes, variables ranging from 20 to over 3,000,000, and more than 2,600 benchmark submissions from 24 contributing institutions at launch.

The core provocation embedded in QOBLIB's design is already producing results that should temper near-term optimism: since the library's initial preprint release, improvements to classical solvers have nearly doubled the largest solved Market Split instances — from 60 to 110 variables — raising the performance threshold that any quantum processor must clear to claim practical [quantum advantage](https://quantumintel.tech/glossary/quantum-advantage) on those problems. That's the benchmark moving the goalposts, not quantum hardware catching up.

---

## What Is QOBLIB and Why Does It Matter?

The "Intractable Decathlon" — the consortium's informal name for its ten NP-hard combinatorial problem classes — addresses a structural credibility problem in quantum optimization. Algorithms like [QAOA](https://quantumintel.tech/glossary/qaoa) and quantum annealing carry no theoretical performance guarantees. Every advantage claim made to date has been empirically contested, in part because there was no agreed-upon, hardware-agnostic comparison framework. Researchers could choose favorable problem instances, cherry-pick classical baselines, or benchmark against solvers that were years out of date.

QOBLIB closes those escape routes. The library is explicitly model-independent: problem instances can be formulated as Mixed-Integer Programming (MIP), Integer Linear Programming (ILP), Mixed-Integer Quadratic Programming (MIQP), or QUBO representations, depending on the execution platform. A superconducting processor running a variational circuit, a quantum annealer consuming a QUBO, and a classical MIP solver competing on Gurobi all run the same underlying instances and submit to the same public leaderboard.

The ten problem classes span instance sizes from tens to tens of thousands of decision variables — the regime where [NISQ](https://quantumintel.tech/glossary/nisq) hardware is actually being deployed today. Critically, the library was designed to scale: as hardware improves, harder instances are already in the repository waiting.

Lead authors are Thorsten Koch of ZIB and TU Berlin and Stefan Woerner of IBM Quantum, with co-authors drawn across the full founding consortium.

---

## The Leaderboard Architecture

QOBLIB's public web portal provides three tools that distinguish it from prior benchmarking efforts:

1. **Interactive complexity-landscape visualization** — maps all 1,264 instances by problem size and matrix density, giving researchers a visual map of where classical solvers break down.
2. **Live best-known solution registry** — tracks both classical and quantum best solutions in real time, so any claimed improvement is immediately visible and contestable by the community.
3. **Automated submission builder** — validates pull requests programmatically to enforce reproducible, fair reporting. This is the mechanism that prevents the selective disclosure that has historically plagued quantum performance claims.

At launch, over 2,600 submissions have already been logged, a signal that the community was waiting for exactly this infrastructure.

---

## Who's Participating — and What That Reveals

The contributor list at launch is a useful map of where serious quantum optimization work is actually happening. It spans:

- **Industrial end-users:** E.ON Digital Technology (energy sector optimization)
- **National supercomputing centers:** STFC Hartree Centre, Forschungszentrum Jülich
- **Academic groups:** National University of Singapore, University of Southern California, City University of Hong Kong
- **Commercial quantum software providers:** Q-CTRL, Kipu Quantum, Aqarios, Qunova Computing, ParityQC, JIJ, and Qoro Quantum

The diversity matters. Prior benchmarking initiatives have tended to be hardware-vendor-led, which introduces obvious incentive problems. QOBLIB's governance structure — academic-led, with IBM Quantum as a co-author rather than sole proprietor — provides more credibility. That said, IBM Quantum's fingerprints are visible in the architecture, and independent replication of the submission validation pipeline will be important for long-term trust.

The inclusion of E.ON Digital Technology signals that at least one large industrial buyer is engaged seriously enough to contribute data, not just consume it. For enterprise buyers evaluating quantum optimization vendors, that participation is worth noting: real procurement decisions may eventually reference QOBLIB scores.

---

## The Classical Baseline Problem Is More Acute Than Most Acknowledge

The Market Split data point buried in the source material deserves emphasis. Between QOBLIB's preprint and its formal *Nature Computational Science* publication, classical solvers — without any quantum assistance — nearly doubled the scale of problems they could solve in that class (60 to 110 variables). This is not an isolated phenomenon. It reflects a pattern that has repeated across quantum optimization benchmarking efforts for years: quantum hardware roadmaps improve, but so do classical algorithms, often faster.

For investors and enterprise buyers, the practical implication is straightforward: any vendor claiming quantum advantage on optimization should be asked which classical baseline they used, and when that baseline was last updated. QOBLIB now provides the infrastructure to make that question answerable with a public, auditable record rather than a vendor's internal testing.

---

## Industry Trajectory

QOBLIB's publication in *Nature Computational Science* gives it peer-reviewed legitimacy that prior benchmarking proposals lacked. The open-source repository and automated submission system lower the barrier for hardware vendors, software providers, and end-users to participate without requiring trust in a single gatekeeper.

The library's design also anticipates fault-tolerant hardware. With problem instances extending to over 3,000,000 variables, QOBLIB has structured room for [fault-tolerant quantum computing](https://quantumintel.tech/glossary/fault-tolerant-quantum-computing) systems — should they arrive — to compete on the same leaderboard alongside today's NISQ devices. That continuity of measurement across hardware generations is one of QOBLIB's most practically valuable properties.

For the quantum optimization software ecosystem, QOBLIB creates a new competitive dynamic. Companies like Kipu Quantum, Q-CTRL, and ParityQC that contributed at launch are now publicly accountable to a shared standard. Future marketing claims that can't be grounded in QOBLIB submissions will face harder scrutiny from technically informed buyers.

---

## Key Takeaways

- **QOBLIB** is an open-source, peer-reviewed benchmarking library published in *Nature Computational Science*, led by IBM Quantum, ZIB, TU Berlin, and Purdue University.
- The library contains **1,264 curated problem instances** across ten NP-hard combinatorial optimization classes, with variables ranging from 20 to over 3,000,000.
- At launch, **2,600+ submissions** have been logged across **24 contributing institutions**, including industrial users, national supercomputing centers, and commercial quantum software vendors.
- Since the preprint release, classical solver improvements **nearly doubled** the largest solved Market Split instances (60 → 110 variables), illustrating the moving-target problem for quantum advantage claims.
- QOBLIB is **model-independent**: instances can be formulated as MIP, ILP, MIQP, or QUBO, enabling fair cross-paradigm comparison between quantum and classical approaches.
- The automated submission builder enforces reproducible reporting — addressing a known credibility problem in prior quantum benchmarking.

---

## Frequently Asked Questions

**What is QOBLIB?**
QOBLIB (Quantum Optimization Benchmarking Library) is an open-source benchmarking framework published in *Nature Computational Science* by a consortium including IBM Quantum, Zuse Institute Berlin, TU Berlin, and Purdue University. It provides 1,264 standardized NP-hard problem instances and a public leaderboard for comparing quantum, classical, and hybrid optimization algorithms on equal footing.

**How does QOBLIB differ from previous quantum benchmarks?**
Unlike hardware-specific benchmarks, QOBLIB is model-independent — the same problem instances can be expressed as MIP, ILP, MIQP, or QUBO formulations, allowing comparison across radically different computational paradigms. It also enforces reproducibility through an automated submission builder and publicly tracks best-known solutions in real time.

**Has any quantum algorithm demonstrated advantage on QOBLIB instances?**
The source material does not report quantum advantage on any QOBLIB instance. More notably, since the preprint's release, classical solvers improved enough to nearly double the largest solved Market Split instances, raising the bar quantum hardware must clear.

**Who can submit results to QOBLIB?**
The library uses an automated pull-request validation system, making it open to any researcher or organization. At launch, 24 institutions had already submitted, spanning academia, national labs, industrial end-users, and commercial quantum software providers.

**Why does QOBLIB matter for enterprise buyers evaluating quantum optimization vendors?**
It provides a vendor-neutral, auditable performance record. Buyers can now ask vendors to demonstrate performance on standardized QOBLIB instances rather than accepting proprietary benchmark claims. The live leaderboard means any advantage claim can be independently verified — or refuted — against current classical baselines.