Five mistakes in a million tries. That's the new bar for quantum computing accuracy, and it was set on a machine customers can already use commercially.

  • 99.9995% logical fidelity achieved on Quantinuum's Helios computer
  • 10x lower error rate than Quantinuum's own record from just last month
  • 100x lower error rate than the best published result from any other hardware type
  • 2029 target year for Apollo, Quantinuum's planned fully fault-tolerant quantum computer

What "five errors in a million" means

On October 19, 2026, Quantinuum announced a world-record quantum error-correction result: a logical fidelity of 99.9995% on its Helios commercial quantum computer, using the company's own error-correction architecture.

Quantum bits, or qubits, are notoriously fragile — tiny disturbances from heat, stray electromagnetic fields, or simple hardware imperfection can flip a calculation from right to wrong. Error correction bundles many fragile physical qubits into a single sturdier "logical" qubit, and fidelity measures how often that logical qubit gets the right answer. A rate of 99.9995% works out to roughly five errors for every million operations — a threshold squarely in the range needed for practical, large-scale quantum computing.

How Quantinuum got there twice in two months

This is Quantinuum's second error-correction breakthrough in as many months. In September 2026, the company set its previous record; the new result cuts that error rate by a further 10 times, using a different error-correction code on the same Helios hardware.

Quantinuum's architecture combines tailored AI-driven decoders, all-to-all qubit connectivity, and high-fidelity trapped-ion operations, letting engineers test and swap between different error-correction codes on the same machine rather than redesigning hardware for each attempt. In the latest run, the team used a well-known "color code" to reach the record figure — one that comes out 100 times more accurate than the best published result from a superconducting system, the rival hardware approach used by companies including Google.

"We have reduced error rates tenfold in rapid succession using a different code on the same commercial hardware, demonstrating the flexibility of our error-correction architecture," said Dr. Rajeeb Hazra, CEO of Quantinuum. "We've crossed the threshold necessary for exponential improvements in accuracy, reinforcing our path to Apollo and how we can help users unlock commercial value from quantum at scale."

The road to Apollo

Quantinuum says the accuracy demonstrated on Helios is now within the target range for Apollo, its planned universal, fully fault-tolerant quantum computer slated for 2029. Fault tolerance is widely viewed as the dividing line between today's experimental quantum machines and systems capable of reliably outperforming classical computers on real-world problems — in drug discovery, materials science, and cryptography among other fields.

Rapid, repeatable error-rate improvements on hardware that customers can already rent commercially are a different kind of milestone than a one-off lab demonstration. They suggest the gap between "quantum computer that works in principle" and "quantum computer that works at scale" is closing faster than many in the field expected even a year ago — five errors in a million at a time.