TL;DR: Quantinuum's Helios trapped-ion computer held a logical qubit to 99.9995% fidelity — five errors in a million operations — on October 19, 2026.
A record logical memory on commercial hardware
Quantinuum announced on October 19, 2026, in Broomfield, Colo., that its commercial Helios computer achieved a logical memory fidelity of 99.9995% using the company's error-correction architecture. According to the company's announcement titled "Quantinuum Closes In on One Million Error-Free Operations," the result equals five uncorrectable errors per million operations. The company described the outcome as a world record for logical accuracy through quantum error correction.
The paper reporting the work was posted October 19 on the company's website and arXiv.org, according to coverage in Science News titled "A quantum computing feat brings the ‘megaquops’ era a step closer." The paper is listed as A. Lavasani et al., Quantum error correction well below the color code threshold, arXiv.org, 2026, doi: 10.48550/arXiv.2610.24077. Quantinuum identified Helios as commercial trapped-ion hardware.
Quantinuum said the error rate is ten times lower than its September 2026 record and more than 100 times lower than the superconducting record. The company attributed the superconducting comparison point to Google Quantum AI using a superconducting system. Quantinuum also said the achievement was its second breakthrough in two months on Helios.
Why five in a million matters for megaquops
Researchers have described a "megaquop" machine as a quantum computer that can run over a million operations error-free, according to Quantinuum. Science News reported that quantum computers capable of doing a million reliable operations could bring a new stage of quantum calculation. Quantinuum said its latest result places it on the cusp of that achievement.
David Hayes, a quantum physicist at Quantinuum based in Broomfield, Colo., said the statistics are consistent with the result being a good building block for a megaquop machine, according to Science News. Hayes also noted that reaching that regime will require scaling up the computers. The company said scaling toward longer, more reliable quantum computations requires exponential error suppression.
The October experiment involved only a single logical qubit, according to Science News. Reaching the megaquop regime could possibly involve around 50 to 100 logical qubits, the publication reported. Quantinuum said an Apollo-scale machine could reach the fault-tolerant quantum advantage regime using 30 times fewer physical qubits than other leading approaches.
How distance scaling cut errors by twelvefold steps
The result was enabled by distance scaling, the prediction that error rates should be suppressed exponentially as longer-distance codes using more physical qubits per logical qubit are used, according to Quantinuum. In experiments using color codes of distance 3, 5 and 7, Helios reduced the logical failure rate by an average factor of 12 each time the code distance increased. The company reported the work was performed with code written in Guppy.
At distance 7, the system achieved logical fidelity of 99.9995% per syndrome-extraction cycle, equivalent to a failure probability of just 5 × 10⁻⁶, according to Quantinuum. The company described that outcome as nearly "six nines." Experimental data and simulation results showing logical error per round at code distances 3, 5 and 7 were presented with slope Lambda indicating suppression by a factor of 12 for each step in code distance.
To measure a failure rate that rarely fails, the team ran the memory for a long time to average down uncertainty, according to Quantinuum. The company said that when an error occurs five times in a million rounds, extended averaging is needed to observe even two errors with reasonable statistics. Adam Reed, lead physicist for commercial operations, said the team felt excitement and suspense as results arrived because each rare error mattered.
Science News reported that a logical qubit composed of 37 physical qubits had just five errors in a million operations. Quantinuum said the color code used up to 37 physical qubits per logical qubit. Helios itself is made of 98 quantum bits, or qubits, according to Science News.
Trapped ions and the logical-qubit approach
Helios is made of electrically charged atoms, called ions, hovering over a chip, according to Science News. Those ions serve as the qubits of the computer. Quantinuum described its system as trapped-ion technology with high-fidelity physical operations and all-to-all qubit connectivity.
Like all qubits, those ions are finicky, with delicate states that can flip unintentionally and disrupt calculations, such as a 0 randomly switching to a 1, according to Science News. Quantum error correction addresses this by combining multiple error-prone qubits into one unit called a logical qubit that is less prone to mistakes. Quantinuum described its approach as an end-to-end logical error-correction architecture programmable through its developer platform.
The company said the architecture integrates tailored AI decoders, all-to-all qubit connectivity, high-fidelity physical operations and hardware-level error correction. It also allows testing, refining and combining different error-correction codes to optimize accuracy and efficiency. Quantinuum said the architecture allows adaptive syndrome extraction, which adjusts error-checking measurements during computation, and chain-mapping, which connects codes suited to different computational tasks.
Quantinuum said it had already shown quantum memories lasting over 20 seconds. It described the latest demonstration as the most robust quantum memory ever recorded. The company also said it had previously demonstrated numerous quantum error-correction codes, primitives needed for logical computation, steadily declining logical error rates and full computations at the logical level.
Lambda 12 versus Google's lambda 2
Logical qubits come in different sizes, integrating a given number of physical qubits, according to Science News. Each time a logical qubit is scaled up, the error rate drops by a factor called lambda. Science News reported that a previous demonstration by Google, reported in Nature in 2024, had a lambda of about 2, meaning errors were cut in half each time logical qubits were scaled up.
Quantinuum's computer has a lambda of about 10, according to Science News. The company itself reported an average factor of 12 for logical memory errors. The company also reported a suppression factor Lambda of 9 for logical gate errors when stepping code distance from 3 to 5.
Ken Brown, a quantum engineer at Duke University and a scientific adviser for Logiqal who was not involved with the research, said the large lambda represents a major savings in the overall number of qubits needed, according to Science News. Brown said achieving a desired error rate takes fewer physical qubits when lambda is large. Quantinuum said it believes the result suggests fault tolerance can be delivered with significantly fewer qubits than required for other modalities.
Robin Blume-Kohout, co-director of Sandia National Laboratories' Quantum Performance Lab, who was not involved in the study, said the paper demonstrates real performance gains from quantum error correction, according to Quantinuum. Blume-Kohout said the logical qubit was nearly 100 times better than its component physical qubits. He said a scaled-up version could reach reliability required for useful fault-tolerant computation at less than a million physical qubits.
Gates, Helix code and the path to Apollo
The team also measured two-qubit Clifford gate fidelity at increasing code distance, according to Quantinuum. The company measured a suppression factor Lambda of 9 when stepping code distance from 3 to 5. It said the result again demonstrated exponential error suppression on its hardware.
Quantinuum described three crucial components for a fully fault-tolerant processor built with this code under a Clifford plus T model with T gates teleported via magic-state injection: magic-state preparation, two-qubit Clifford gates and syndrome-extraction gadgets. The company said the latest work demonstrates high-accuracy implementation and below-threshold scaling of the latter two. It said magic-state preparation with the same code was demonstrated in a separate paper in 2025, while real-time decoding was first demonstrated in 2021 and has since been refined.
The company said nobody has yet demonstrated below-threshold scaling for magic-state preparation. It said almost everything had already been put together in the Helix code and that integration of all elements would follow. In September 2026, Quantinuum used the same error-correction architecture with its Helix code to demonstrate record-setting logical memory and computation, according to the company. Both studies ran on the same Helios hardware.
On Helios, Helix used 10 physical qubits per logical qubit, compared with up to 37 for the color code, while the color code delivered lower logical error rates, according to Quantinuum. The company said the comparison provides insight into which applications pair best with each code as systems advance toward fault tolerance. It said successive breakthroughs using different codes on the same commercial hardware show how accuracy and qubit efficiency can be optimized.
Rajeeb Hazra, chief executive of Quantinuum, said the result proved the company has the error-correction architecture and trapped-ion technology required for universal fault tolerance, according to the company announcement. Hazra said error rates were reduced tenfold in rapid succession using a different code on the same commercial hardware. Quantinuum said its accuracy performance is now within the target range for Apollo, its planned universal fault-tolerant quantum computer expected to be available by 2029.
Quantinuum said it is combining hardware, error-correction architecture and full-stack software into an integrated system for fault-tolerant quantum computing. It said it is providing tools for developing quantum applications addressing materials science, chemistry, finance, logistics and other sectors. The full research was made available through quantinuum.com/fault-tolerance-by-2029.