IBM and researchers at the University of Chicago have run a quantum computation on 70 error-corrected logical qubits that finished in about 15 minutes, a job the team says leading classical simulation methods cannot reproduce at any practical cost.

The work appears in "Sampling hard circuits with verifiably high fidelity," posted to arXiv on 28 July 2026. The paper documents one of the largest logical-qubit demonstrations reported so far.

What did the system actually compute?

The experiment built on random circuit sampling, a long-standing benchmark that asks a quantum machine to produce output patterns too tangled for classical hardware to replicate efficiently.

The team replaced that benchmark with a more structured variant. The new circuit kept the same classical-hardness properties, while adding enough internal order for the researchers to detect errors during the run itself.

That combination, harder than random sampling yet still checkable, sits at the heart of the claim.

How did the hardware behave?

The researchers encoded 70 logical qubits and pushed them through 2,415 logical two-qubit operations and 468 logical T gates. Effective logical error rates came in at roughly one-tenth of the underlying physical-qubit error rate, an order-of-magnitude improvement that the group attributes to the encoding scheme.

Running the encoded circuit took about 15 minutes. Reproducing the same calculation with leading classical simulation methods would require prohibitive runtimes, according to the paper.

"Verification remains one of the biggest challenges in firmly establishing experimental quantum advantage," said Bill Fefferman, Associate Professor at the University of Chicago. "This experiment develops techniques to better characterize the fidelity of hard quantum states under noise, increasing confidence that the quantum computer is solving a computationally hard problem."

PhD student Soumik Ghosh, also at Chicago, framed the result more broadly. "Beyond strengthening experimental validation, advances in verification have the potential to unlock practical applications for the next generation of quantum computers," he said.

Where does this sit on the roadmap?

IBM has staked a public position on what the run means. "We are now firmly in the quantum advantage era," said Jay Gambetta, Director of IBM Research and IBM Fellow. "We have demonstrated a quantum computation beyond the practical reach of classical computers that establishes, with statistical confidence, a lower bound on how faithfully it was executed."

That is a marketing sentence dressed in measurement language. Readers can judge the substance on its own: 70 logical qubits, a 15-minute wall-clock time, and a verification scheme that survived peer review at arXiv.

The Quantum Advantage Tracker, maintained by the research community, now lists the experiment among the publicly logged demonstrations.

What changes for the next machine?

Two things will determine whether the next IBM system repeats the trick at scale. The first is logical-qubit count. Hitting 145 logical qubits, roughly twice the size reported here, would push the classical gap wider by another order of magnitude. The second is whether the verification protocol travels with the system, rather than collapsing once the circuit grows denser.

The paper carries DOI 10.48550/arXiv.2607.25941 and lists Simon Martiel, Jay-U Chung, Alireza Seif, Soumik Ghosh, Ian Hincks, Abhinav Deshpande, Bill Fefferman, Jay M. Gambetta, and Ali Javadi-Abhari as authors. The circuits and experimental results have been released publicly through the Quantum Advantage Tracker, a deliberate move that lets outside groups attempt independent checks.

The next test will arrive when a larger logical-qubit machine attempts a harder problem under the same protocol, and the field can see whether IBM's claim of a firmly arrived advantage era holds at the next integer.