Types of quantum computer, compared

Quantum computers are built in very different ways. Each approach has strengths and weaknesses that show up in the benchmarks. This page describes the patterns; it doesn’t pick a winner.

Superconducting

Tiny electrical circuits on a chip, cooled to near absolute zero. Built by Google, IBM, Rigetti and IQM, among others. They run operations very fast, but each qubit only connects to its neighbours on the chip.

In this release: Rigetti Computing Ankaa-3, Rigetti Computing Aspen-M-1, Rigetti Computing Cepheus-1-108Q, IQM Quantum Computers Emerald, IQM Quantum Computers Garnet, IBM Heron r2, IBM Heron r3, IBM Nighthawk r2, Google Willow.

Trapped ion

Individual charged atoms held in place by electric fields and controlled with lasers or microwaves. Built by Quantinuum and IonQ, among others. They tend to make fewer mistakes per operation and any qubit can interact with any other, but they run more slowly.

In this release: IonQ Forte, Quantinuum H1, Quantinuum H2, IonQ Harmony, Quantinuum Helios, IonQ Oxford Ionics two-qubit prototype, IonQ Tempo.

Neutral atom

Uncharged atoms held in grids of laser light (optical tweezers) that can be moved around. Built by QuEra, Atom Computing, Pasqal and Infleqtion, among others. They can hold very large numbers of atoms and have led several recent error-correction experiments.

In this release: Infleqtion Cesium atom array, QuEra Computing Harvard–MIT rubidium atom array, Atom Computing Ytterbium atom array.

Quantum annealing

Machines built for one kind of task: finding good answers to optimisation problems. D-Wave makes them. They are not general-purpose gate-based computers, so their results are compared only with each other.

In this release: D-Wave 2000Q, D-Wave Advantage.

What the results show so far

In the first QUOPS results, trapped-ion machines ran the biggest circuits and superconducting machines ran them far faster (see the results). Gate errors are lowest on trapped-ion machines, at around one in a thousand (see the results). These are patterns in the published numbers, not verdicts: each design is improving quickly.