Record in Numbers
On October 19, 2026, Quantinuum announced a world record in quantum error correction: its commercial Helios computer achieved 99.9995% logical fidelity (company announcement). That is just five errors per million operations. Logical fidelity is the probability that a qubit that has undergone error correction returns the correct result at the end of a computation; 99.9995% denotes precisely that level of accuracy. A commercial computer means a system built for customers, not a prototype in a research lab.
The error rate is ten times lower than the company's own record from September.
That accuracy matters in practice: in fault-tolerant computing, thousands of logical operations run in sequence, and the reliability of each operation determines the final result. Every reduction in logical error therefore increases the volume of computation a system can perform. Compared with the best published result on any other hardware platform — Google Quantum AI's superconducting system — the gap reaches 100x.
According to the company's statement, the result confirms its leading position on the path to fault-tolerant quantum computing. The record was set within an error-correction architecture on trapped-ion hardware using a method known as the color code — one of the well-established approaches in the field.
How the Errors Were Suppressed
The biggest obstacle in quantum computing is the fragility of qubits: even the slightest external disturbance can corrupt a computation's result. The idea behind error correction is simple: several physical qubits are combined into one logical qubit, and redundancy is used to detect and correct errors. The difficulty is that each additional physical qubit is itself a source of errors — so a system's real gain is measured at the point where the logical qubit starts making fewer errors than the physical qubits it is made of; in the industry this threshold is called breakeven.
In trapped-ion technology, individual atoms are held in place by an electromagnetic field, and their quantum states are used for computation. In Quantinuum's experiment, as the code distance — the parameter that sets how many physical qubits are enlisted to protect quantum information — increased, logical errors fell exponentially: at each step, the remaining errors shrank by a record factor of 12.
The color code belongs to the family of topological error-correction codes: qubits are arranged as a geometric lattice, and errors are detected using the lattice's symmetry. The advantage of such codes is that, when the hardware's physical errors are below a certain threshold, increasing the code size reduces error exponentially — exactly the regime demonstrated in Quantinuum's experiment.
This metric matters because it is precisely this exponential suppression that makes it possible to scale fault-tolerant systems: the larger the code distance, the more errors the system can correct and the smaller the final error — but at the cost of more physical qubits. As code distance grows, the number of additional physical qubits grows too, so efficiency is measured precisely by the error reduction gained per added qubit.
Road to Apollo: 2029
CEO Rajeeb Hazra commented on the result in an interview with The Wall Street Journal:
"We have the world's lowest error rate and the most efficient path to achieving it."
According to Hazra, this breakthrough brings the company closer to Project Apollo — the name of the universal fault-tolerant quantum computer expected in 2029. Universal means a system capable of performing various computational tasks, not one built for a single specialized problem. Fault-tolerant means the system can independently detect and correct its own errors. He added that Apollo will operate with roughly 30x fewer physical qubits than other leading approaches.
What Independent Experts Say
Robin Blume-Kohout, co-lead of the Quantum Performance Laboratory at Sandia National Laboratories, who was not involved in the study, called the result a significant milestone that industry researchers have pursued for many years. By his estimate, the logical qubit in the experiment performed nearly 100x better than the physical qubits it was composed of — direct evidence that error correction is delivering serious practical benefit. The fact that the evaluator did not participate in the study makes his assessment independent.
Such independent assessment matters for scientific results: a positive evaluation from an expert outside the study's authors increases confidence in the result.
Ars Technica also confirmed the result's significance in its October 20 review: the demonstration showed that errors in error-corrected qubits fall well below the breakeven threshold as the code grows.
