A helium dilution refrigerator photographed in a laboratory, with its stacked metal stages

· Translated from the Spanish original

Quantum computing

Quantinuum reaches 5 errors per million operations on one logical qubit of its Helios

The company reports 99.9995% logical fidelity with a color code and up to 37 physical qubits. It is a single logical qubit, and its fault-tolerant machine would arrive in 2029.

What happened

  • On October 19, 2026, Quantinuum announced a logical fidelity of 99.9995% on its commercial Helios computer, using its error-correction architecture.
  • It equals about 5 uncorrectable errors per million operations. The company says it is 10 times better than its September mark.
  • The color code used up to 37 physical qubits per logical qubit, against 10 for the earlier Helix code.
  • According to Science News, the result corresponds to a single logical qubit. Quantinuum presents it as validation of its path to a fault-tolerant computer, Apollo, planned for 2029.

Why it matters

  • Quantinuum says that, alongside the figure, errors fell exponentially as the code was strengthened. That is the behavior needed to scale, and it is a company claim.
  • Whoever in Chile safeguards long-lived encrypted data, such as banks, notaries and registries, gets no risk date yet, but does get a signal that the quantum calendar is measured in years, not decades.
  • A single logical qubit is not a useful computer. Hundreds or thousands operating together are needed, and no results are published there.

The number

99.9995%: reported logical fidelity, with up to 37 physical qubits per logical qubit.

Context

The announcement adds to a month of movements in the sector, such as the IQM quantum cloud Bechtle will give the German aerospace center and DARPA’s contract with Xanadu. The comparison with Google Quantum AI, about 100 times better than the published superconducting result, comes from the company and not from a third party.

What’s next

  • Apollo, Quantinuum’s fault-tolerant machine, targets the year 2029.
  • No announced timeline for showing the same performance with several logical qubits at once.
  • An outside replication or evaluation of the figure is still missing.

Bottom line

Error correction is the debate that decides whether quantum arrives, and any published figure is only as good as the number of logical qubits it sustains. Meanwhile, mobile apps still carry no post-quantum code in nearly every case measured.

Sources


Edited by Rodrigo Cornejo. How we select and verify: who writes these notes.

Related notes

← All notes