Stark Quantum Independent. Not a vendor.

Briefs

Matrix-product states simulate a 15-to-1 distillation circuit in under 40 seconds

A matrix-product-state simulator ran a 15-to-1 magic state distillation circuit with hundreds of error-correction rounds using 11 logical qubits and 187 physical qubits, at a maximal bond dimension of 64, in under 40 seconds, alongside rotated surface code memory to distance 11 and logical Bell-state preparation to distance 9. The authors attribute several orders of magnitude of that gain to implementation choices rather than to the method, which is the useful part: the cost of simulating these circuits is not fixed, and published resource estimates inherit whatever assumption was made about it. None of this ran on hardware. It is a classical simulation of the circuits a fault-tolerant machine would have to execute, so it constrains resource estimates rather than saying anything about a device.

The numbers

  • 15-to-1 magic state distillation with hundreds of QEC rounds, simulated on 11 logical qubits (187 physical qubits) at maximal bond dimension 64, in under 40 seconds
  • Rotated surface code memory simulated up to distance 11
  • Logical Bell-state preparation simulated up to distance 9
  • Bond dimensions and simulation time reduced by several orders of magnitude against a naive implementation

What it does not show

  • Classical simulation throughout. No part of this ran on quantum hardware.
  • The speed comes from implementation choices the authors optimised, not from the method alone; a naive matrix-product-state implementation is reported orders of magnitude slower.
  • The authors position the method as a complement to near-Clifford simulators, not a replacement for them.

Primary source

arXiv:2609.03073 - Pineiro Orioli et al., Optimized Matrix-Product State Simulations of Quantum Error Correction Circuits

Briefs are frozen after publication. If something here is wrong, the correction appears as a dated note at the top of this page rather than as a silent edit.