Quantinuum and Sandia National Laboratories have put a very specific claim on the record: the 98-qubit Helios system is now accurate enough to push beyond what classical simulation can comfortably track, while keeping error rates low enough to matter. The headline number is less about raw qubit count than about control, with single-qubit operations reaching 99.9975% accuracy and two-qubit operations hitting 99.921%.
That sounds like a lab brag until you remember how unforgiving quantum hardware still is. One misbehaving laser, one tiny ion shift, and the whole calculation can go sideways; the fact that Helios can sustain this level of precision is a sign that the race is moving from ”can it work at all?” toward ”can it be engineered reliably?”
Why 98 qubits matter less than the error rate
Helios is a commercial trapped-ion processor, and that architecture gives it an advantage in control even if it comes with serious infrastructure baggage. Ions are held in electromagnetic fields and moved between processing zones, which makes them easier to manage than some alternative designs, but also demands tight cooling and measurement systems to keep the platform stable as it grows.
Sandia’s researchers have spent more than 20 years working on quantum computing programs under a cooperative research and development agreement with Quantinuum, so this was not a one-off demo cooked up for a press release. The broader point is familiar across the industry: scaling qubits is easy to talk about, but scaling trustworthy qubits is what separates the contenders from the also-rans.
Mid-circuit measurements are the real test
A big part of the work focused on ”mid-circuit measurements,” a mouthful for the ability to measure some qubits during a computation without collapsing everything. That is the sort of plumbing quantum error correction depends on, and it is also where many systems get exposed as more fragile than their marketing suggests.
Sandia built custom benchmarking methods for those operations, which is sensible because standard test suites can miss the parts that matter most for fault-tolerant computing. In other words, the industry is past the stage where a high qubit count alone impresses anyone serious.
- Single-qubit operation accuracy: 99.9975%
- Two-qubit operation accuracy: 99.921%
- System size: 98 qubits
- Architecture: trapped ions
Quantum hardware is shifting toward engineering discipline
Sandia’s quantum work sits inside a long-running DOE-linked effort that has implications for cryptography, pharmaceuticals, energy, and communications, which is a reminder that governments are not funding this just for the fun of it. Meanwhile, companies across the field are chasing different scaling paths: some lean on superconducting chips, while others are betting on optics and hybrid approaches to get more stable systems with lower energy use.
Sandia is also working on integrated photonics, which could help move quantum information with light through microscopic optical structures. If that line of research keeps advancing alongside trapped-ion platforms like Helios, the next milestone will not be a single spectacular benchmark but a repeatable machine that stays accurate as it gets larger.
For now, the interesting question is whether Helios marks the point where quantum computing becomes less of a physics curiosity and more of an engineering contest. The answer will not come from a bigger qubit number alone; it will come from which team can keep the errors boring, predictable, and small.