IBM and research partners have demonstrated quantum processors performing tasks that exceed the reach of leading classical supercomputers, using new verification methods to establish trust in quantum outputs where classical checks fail
IBM, in collaboration with academic and industry partners, has reported a series of experiments in which quantum processors executed computational tasks beyond the practical capabilities of current supercomputers. The demonstrations, conducted with the University of Chicago, Qedma Quantum Computing, Algorithmiq, RIKEN, and BlueQubit, focused on both the scale of the quantum circuits and the reliability of their outputs. Each experiment introduced a distinct verification framework, addressing the challenge of certifying quantum results when classical simulation becomes infeasible.
Logical Qubits and Spacetime Codes
In partnership with the University of Chicago, IBM researchers implemented a sampling task using 70 logical qubits on their Heron quantum processor. The experiment used doped Clifford circuits, a structured alternative to random circuit sampling, embedding non-Clifford gates within a classically simulable reference circuit. This approach allowed the team to benchmark the quantum device against a tractable baseline before extending into a regime where classical simulation is intractable. The circuits incorporated 2,415 logical two-qubit gates and 468 logical T gates, with logical encoding reducing the effective gate error rate by a factor of ten compared to the underlying physical hardware. The quantum processor completed the sampling task in approximately 15 minutes, while leading classical methods faced prohibitive runtimes. Instead of relying on statistical proxies, the experiment used spacetime code techniques to establish a mathematically rigorous lower bound on logical fidelity, providing a self-certifying measure of output quality.
Quantum Simulation of Floquet Dynamics
Qedma Quantum Computing, working with IBM and RIKEN, used IBM's Heron processors to simulate the long-time dynamics of a two-dimensional Floquet Ising model-a system relevant to optoelectronics and light-induced superconductivity. The experiment scaled up to 74 physical qubits and employed Qedma's QESEM software for error suppression and mitigation. Classical simulations were performed on the Fugaku supercomputer using tensor-network and other advanced algorithms, but as the system size and simulation time increased, classical methods diverged and failed to produce consistent results. The quantum processor, with error mitigation, resolved persistent oscillatory behavior that classical approaches could not capture. To further validate the results, core circuit segments were independently reproduced on Quantinuum's trapped-ion hardware, and both heuristic and provable error-mitigation estimators were applied.
Process Validation in Disordered Quantum Materials
Algorithmiq, based in Helsinki and Milan, collaborated with IBM to simulate the operator Loschmidt echo-a measure of information propagation in heterogeneous quantum materials-using 56-qubit circuits on Heron hardware. In this regime, three independent classical simulation groups produced conflicting predictions, highlighting the absence of a definitive classical baseline. To address this, Algorithmiq varied hardware noise profiles and gate calibrations across five IBM quantum processors, demonstrating output stability under controlled noise manipulation. The team also released an open-source classical simulation tool, monoprop, to support transparent benchmarking. This approach emphasizes process-based trust in quantum outputs when exact classical verification is not possible.
Verification and the Limits of Classical Comparison
Across all three demonstrations, the central technical advance was not only the execution of large-scale quantum circuits, but the development of verification strategies that remain meaningful when classical simulation fails. These included logical error bounds derived from spacetime codes, hardware-agnostic error mitigation, and systematic noise manipulation. The experiments were published alongside public circuit repositories, enabling ongoing community benchmarking. The shift toward verifiable quantum outputs is seen as a necessary step for transitioning quantum computing from theoretical promise to practical utility in fields such as drug discovery and materials science. For context, similar efforts to benchmark quantum hardware against classical simulation have been reported in other platforms, such as the deployment of digital twins for neutral-atom quantum processors described in recent Science Report coverage.
Understanding the distinction between physical and logical qubits is essential for interpreting these results. A physical qubit is a single controllable quantum system, such as a superconducting circuit or trapped ion, while a logical qubit encodes information across multiple physical qubits using error-correcting codes. Logical qubits are designed to detect and correct errors, reducing the impact of noise and hardware imperfections. Demonstrating reliable logical-qubit operation at scale is a key milestone toward fault-tolerant quantum computing, but it requires not only high-fidelity hardware, but also robust verification methods-especially as quantum circuits grow beyond the reach of classical simulation.
