State of Quantum Hardware — Week 29, 2026

Published 2026-07-15 · auto-generated from the public corpus · subscribe

This week's fleet

The Provenova corpus holds 28 device snapshots drawn from six hardware providers this week: IBM contributes 17 snapshots, IQM and Origin Quantum three each, Quantinuum and IonQ two each, and Rigetti one. The most recent corpus refresh attempt ran on 2026-07-15. Rankings below come from two source classes, and the distinction governs how each should be read: vendor-reported specifications are manufacturer claims that Provenova does not independently verify, while Metriq submissions are community benchmark results published under CC-BY-4.0.

Two-qubit gate fidelity (vendor-reported). On the two-qubit fidelity board, Quantinuum's H2 ranks first at 0.999 across 56 qubits, captured 2026-07-03, and carries a vendor-reported Quantum Volume of 33,554,432 (2^25). IonQ's Forte follows at 0.996 (36 qubits, 36 algorithmic qubits) and IQM's Garnet at 0.9951 (20 qubits). Rigetti's Ankaa-3 reports 0.995 over 84 qubits, but its figure was captured on 2024-12-23 — the oldest data point in the group by more than eighteen months — so a strict ranking here compares devices measured at very different points in their calibration history. IonQ's Aria completes the five at 0.994 (25 qubits, 25 algorithmic qubits). Every entry on this board is vendor-reported, sourced from quantinuum.com, ionq.com, rigetti.com and the AWS Braket listing for Garnet, so all five values are best-case specifications rather than measured, redistributable results.

Error per layered gate (EPLG, Metriq). EPLG accumulates error per circuit layer, so lower is better. On the EPLG board, IBM's ibm_boston ranks first at 0.00154 (156 qubits, captured 2026-02-18), ahead of Quantinuum's h2-2 at 0.00180 (56 qubits). ibm_pittsburgh records 0.00273 and ibm_torino 0.00551 (133 qubits), with IQM's iqm_garnet fifth at 0.00823 (20 qubits). Every EPLG figure is a Metriq submission under CC-BY-4.0 and is redistributable.

Circuit layer operations per second (CLOPS, Metriq). CLOPS is a throughput measure where higher is better, and the CLOPS board is entirely IBM this week. ibm_fez records 368,905, followed by ibm_kingston at 362,372, ibm_boston at 360,574, ibm_pittsburgh at 358,613, ibm_marrakesh at 353,642 and ibm_torino at 345,670. All six are 156-qubit devices except ibm_torino at 133 qubits, and all are Metriq-sourced under CC-BY-4.0.

QAOA approximation ratio (Metriq). On the QAOA approximation-ratio board — higher is better — Quantinuum's h2-2 ranks first at 0.8018 (56 qubits), ahead of ibm_boston at 0.6734 and ibm_kingston at 0.6412. ibm_marrakesh (0.5731) and ibm_pittsburgh (0.5695) take the next two places, followed by ibm_torino (0.4560) and ibm_fez (0.4209). The tail is informative: iqm_garnet sits at 0.1432, IQM's iqm_emerald at 0.0099 (with a paired QFT accuracy of 0.1293), and Origin's wukong_72 records a negative ratio of -0.0350 (72 qubits, captured 2026-02-24), meaning its sampled solutions fell below the classical random baseline on that instance. Where a Metriq submission also carries QFT accuracy, h2-2 records 0.967.

Movements

Week 28 was the first "State of Quantum Hardware" edition (published 2026-07-12), which makes this the first week-over-week comparison the platform can draw. The standings are unchanged. The corpus holds the same 28 snapshots with the same provider breakdown as last week (IBM 17, IQM 3, Origin 3, Quantinuum 2, IonQ 2, Rigetti 1), and all four boards — two-qubit fidelity, EPLG, CLOPS and QAOA ratio — carry identical values and ordering to the Week 28 report. The only change on the platform is operational: a corpus refresh attempt ran on 2026-07-15, versus 2026-07-05 for the prior edition, and it did not add or alter any snapshot. No captured-at date advanced, so no device re-benchmarked into or out of a board this week.

New on the platform

No new research cards were published during ISO week 29: the card counter reports zero created today and none dated within this week. The corpus holds four research cards in total, and the most recent two — dated 2026-07-11, within Week 28 — remain the newest additions:

Each card pairs a published circuit with its arXiv attribution and a deterministic simulator run; the platform's daily publication cap is three cards.

Method note

Provenova aggregates figures by source class, and every value above is tagged by its origin. Vendor-reported specifications (the two-qubit fidelity board) are manufacturer claims taken from provider material — quantinuum.com, ionq.com, rigetti.com and the AWS Braket listing — and are neither independently verified nor redistributable as raw data. Metriq submissions (the EPLG, CLOPS and QAOA-ratio boards) are community benchmarks published under CC-BY-4.0 and are redistributable. IBM device calibration enters the corpus under Apache-2.0. All platform algorithm runs behind the research cards are deterministic simulator executions rather than hardware executions, so a card's result reflects the compiled circuit, not a live device. Captured-at dates span a wide range — from 2024-12-23 for Ankaa-3 to 2026-07-03 for the vendor fidelity figures — and rankings should be read with that spread and each entry's source class in mind. Every number in this report was fetched from the Provenova leaderboard and corpus-status endpoints on 2026-07-15.