Skip to content
Sıfırıncı Dakika
Latest

Quantinuum's quantum computer hits 99.9995% accuracy

Science1 min read

In brief

Quantinuum said its 98-qubit commercial quantum computer Helios achieved 99.9995% logical fidelity, meaning just five errors per million logical operations. According to the company, this cuts its error rate 10x versus its previous record and 100x versus Google Quantum AI's best published result. The result is presented as a step toward Apollo, its planned fault-tolerant system due in 2029.

  • Tested on Helios, a 98-qubit commercial trapped-ion quantum computer
  • 99.9995% logical fidelity, five errors per million operations
  • 10x lower error rate than its previous record, 100x lower than Google's result
  • Used a color code error-correction method
  • Errors dropped 12x with each increase in code distance
  • Targets Apollo, a fault-tolerant system planned for 2029

What happened

Colorado-based Quantinuum announced a new quantum error-correction record on its 98-qubit commercial quantum computer, Helios. According to the company, the system reached 99.9995% logical fidelity, meaning only five errors per million logical operations.

The result cuts the error rate 10x compared with the company's previous record from September 2026 and 100x compared with the best published result from Google Quantum AI.

How error correction works

Quantum computers process data using qubits, which are highly sensitive to outside disturbances such as heat, stray electromagnetic signals, vibrations and cosmic rays. Scientists therefore use quantum error correction, combining multiple physical qubits into more reliable "logical qubits."

For this experiment, Quantinuum used a technique called a color code on Helios. The team also found that each increase in code distance reduced remaining logical errors by a factor of 12.

Two approaches, one machine

A month earlier, the company demonstrated record-setting logical memory and computation using a different method called Helix. Both experiments ran on the same Helios hardware, showing the company can improve performance by changing error-correction methods without replacing the computer. Helix required 10 physical qubits per logical qubit, while the newer color code used as many as 37 but delivered significantly lower error rates.

Why it matters

Quantinuum says the result strengthens its plan to launch Apollo, a universal fault-tolerant quantum computer, by 2029. Robin Blume-Kohout of Sandia National Laboratories, who was not involved in the work, called it an important milestone and said a scaled-up version could reach the reliability needed for useful fault-tolerant computation with fewer than a million physical qubits.

Why it matters

Quantum computers nearing everyday use could fundamentally change computing power across fields including AI; this story shows with a concrete number where that path stands.

Sources

Related stories