Vol. III·Hardware Scorecard·Benchmarks

Every meaningful system · live record

sourced from peer-reviewed papers

The live scorecard.

Continuously updated hardware benchmark across every meaningful quantum system — qubit counts, fidelity, coherence, connectivity.

Verified Provisional Disputed

Field record holders · who currently leads on each metric

Most physical qubits

1,180

gate-based

Atom Computing Phoenix

Highest 2Q fidelity

99.99%

two-qubit gate

IonQ Tempo

Most logical qubits

demonstrated

Quantinuum Helios

Longest coherence

2.8s

T2 — phase stability

Infleqtion Sqale

Fastest gates

25ns

shorter is better

Google Willow

Highest throughput

330K

CLOPS · IBM only

IBM Heron R2 (Kingston/Aachen)

How to read this scorecard

Qubit counts alone are meaningless without fidelity context. A 1,000-qubit machine at 99% gate fidelity does less than a 100-qubit machine at 99.99%.

Logical qubits are not interchangeable — Google's distance-7 surface-code qubit, Quantinuum's Steane-code qubit, and Quantinuum's Iceberg-detected qubit are three different things with error rates differing by orders of magnitude.

"Best pair" vs "median" 2Q fidelity — vendors selectively publish hero numbers. We note the basis. Always prefer median across the chip.

Microsoft Majorana 1 is flagged disputed: peer reviewers, including Nature's own editorial team, have publicly questioned whether the topological qubit claim is supported by the evidence. The underlying 2018 paper was retracted.

Glossary · what each metric actually means⌄

Physical qubits: The raw, noisy qubits in the device. More is generally better, but fidelity matters more than count.
Logical qubits: Error-corrected qubits encoded across many physical qubits. Each costs 100–1000× physical qubits to build, but is orders of magnitude more reliable. The right number to compare for "useful computing."
2Q gate fidelity: Probability that a two-qubit gate executes correctly. The single most important metric. At 99%, you can run ~100 gates before random noise dominates. At 99.99%, ~10,000 gates. The FTQC threshold is ~99.9%.
SPAM fidelity: State Preparation And Measurement. Every shot starts with a small probability of being wrong before any computation. SPAM compounds with gate errors.
T1 / T2: Coherence times. T1 = how long the qubit stays in its energy state (relaxation). T2 = how long the quantum phase stays coherent. T2 is usually the binding constraint. Gate time ÷ T2 = how many ops you can run before decoherence ruins your computation.
Gate time: How long a typical gate takes. Faster is better. Superconducting: ~20–100ns. Trapped ion: ~10–300µs (1000× slower but much higher fidelity).
Connectivity: Which qubits can directly interact. All-to-all (trapped ion) needs no SWAP overhead. Heavy-hex / square (superconducting) need extra SWAPs to move information around.
Quantum Volume / CLOPS / #AQ: Composite throughput metrics. IBM has deprecated QV in favor of layer fidelity. CLOPS = circuit layer operations per second (speed only, not quality). IonQ's #AQ measures algorithmic qubits via QED-C benchmarks.
Confidence flag: Verified = peer-reviewed paper or vendor spec sheet. Provisional = single-source vendor claim, not independently replicated. Disputed = contested by independent analysis (e.g., Microsoft Majorana 1).

Visual map · all systems on one chart

Qubit count × gate fidelity

Each circle is a quantum system. X-axis = number of physical qubits (log scale). Y-axis = two-qubit gate fidelity. Bubble size ∝ logical qubits demonstrated (number shown when ≥5). The teal dashed line marks the ~99.99% fault-tolerance threshold. The corner you want to be in is top-right: high qubit count AND high fidelity. Today, no system is there yet.

Progress tracker

How close to useful fault tolerance?

The journey from a working quantum system to one that actually breaks RSA or runs full chemistry calculations. Each gate below must be cleared. No system has cleared all six — but the leaders have cleared four.

Quantinuum Helios

trapped ion

Progress

4/6

✓
Working system
98 physical qubits
✓
99.9% 2Q fidelity
99.92%
✓
Logical qubit demo
94 logical demo'd
✓
Break-even QEC
Iceberg detection (94); Steane FT (50 below break-even)
5
~100 logical qubits
6
Cryptographically useful
requires ~thousands of logical qubits

Quantinuum H2

trapped ion

Progress

4/6

✓
Working system
56 physical qubits
✓
99.9% 2Q fidelity
99.91%
✓
Logical qubit demo
12 logical demo'd
✓
Break-even QEC
Transversal CNOT with Microsoft (Apr 2026)
5
~100 logical qubits
6
Cryptographically useful
requires ~thousands of logical qubits

Google Willow

superconducting

Progress

3/6

✓
Working system
105 physical qubits
2
99.9% 2Q fidelity
99.88%
✓
Logical qubit demo
1 logical demo'd
✓
Break-even QEC
Surface d=7 (below threshold)
5
~100 logical qubits
6
Cryptographically useful
requires ~thousands of logical qubits

QuEra Gemini

neutral atom

Progress

2/6

✓
Working system
260 physical qubits
2
99.9% 2Q fidelity
99.50%
✓
Logical qubit demo
30 logical demo'd
4
Break-even QEC
Magic state distillation demo
5
~100 logical qubits
6
Cryptographically useful
requires ~thousands of logical qubits

Atom Computing Phoenix

neutral atom

Progress

2/6

✓
Working system
1180 physical qubits
2
99.9% 2Q fidelity
99.50%
✓
Logical qubit demo
28 logical demo'd
4
Break-even QEC
Bernstein-Vazirani demo; 64-logical architecture
5
~100 logical qubits
6
Cryptographically useful
requires ~thousands of logical qubits

Pasqal Orion Gamma

neutral atom

Progress

2/6

✓
Working system
140 physical qubits
2
99.9% 2Q fidelity
99.70%
✓
Logical qubit demo
2 logical demo'd
4
Break-even QEC
5
~100 logical qubits
6
Cryptographically useful
requires ~thousands of logical qubits

IBM Nighthawk

superconducting

Progress

2/6

✓
Working system
120 physical qubits
✓
99.9% 2Q fidelity
99.90%
3
Logical qubit demo
4
Break-even QEC
5
~100 logical qubits
6
Cryptographically useful
requires ~thousands of logical qubits

IonQ Tempo

trapped ion

Progress

2/6

✓
Working system
100 physical qubits
✓
99.9% 2Q fidelity
99.99%
3
Logical qubit demo
4
Break-even QEC
5
~100 logical qubits
6
Cryptographically useful
requires ~thousands of logical qubits

Even the leading systems (Quantinuum Helios, Atom Computing Phoenix, Google Willow) sit at ~4/6. The two remaining gates — 100 logical qubits, then cryptographically useful scale — are the difference between today and ~2030. Real fault tolerance is closer than the headlines suggest, but still meaningfully out.

Per-modality champions

Best system in each technology family

Modality	Most physical qubits	Highest 2Q fidelity	Most logical qubits	Longest T2 coherence
Superconducting	1,121 IBM Condor	99.90% IBM Nighthawk	1 Google Willow	150µs IBM Heron R2 (Kingston/Aachen)
Trapped Ion	100 IonQ Tempo	99.99% IonQ Tempo	94 Quantinuum Helios	1.0s Quantinuum Helios
Neutral Atom	1,180 Atom Computing Phoenix	99.73% Infleqtion Sqale	30 QuEra Gemini	2.8s Infleqtion Sqale
Photonic	12 Quandela Belenos / MosaiQ 12	97.00% Quandela Belenos / MosaiQ 12	12 Xanadu Aurora	— —
Silicon Spin	12 Diraq (imec foundry)	99.64% SQC (donor-based)	— —	1.0s SQC (donor-based)
Annealing	4,400 D-Wave Advantage2	— D-Wave Advantage2	— —	— —
NV-Center	5 Quantum Brilliance QB-QDK2.0	98.00% Quantum Brilliance QB-QDK2.0	— —	1ms Quantum Brilliance QB-QDK2.0

Full data · by modality

Superconducting

System	Physical	Logical	2Q Fid.	1Q Fid.	SPAM	T1	T2	Gate	Connectivity	Source
IBM Heron R2 (Kingston/Aachen) Supports 5,000 2Q-gate circuits at >99.9% layer fidelity on a subset of pairs	156	—	99.70%(med)	99.97%	99.00%	200µs	150µs	68ns	Heavy-hex (degree 3)	link
IBM Nighthawk 5,000 → 7,500 2Q gates by EOY 2026	120	—	99.90%(best)	99.97%	99.00%	350µs	—	70ns	Square lattice, degree 4 (218 couplers)	link
IBM Condor Scaling demo; no QV/median fidelity published for full array	1,121	—	99.00%(med)	99.90%	—	100µs	—	500ns	Heavy-hex	Dec 1, 2024
Google Willow First sub-threshold QEC demonstration on real hardware	105	1	99.88%(med)	99.97%	99.50%	100µs	89µs	25ns	Nearest-neighbor 2D grid (avg degree 3.47)	link
Rigetti Cepheus-1-108Q Target 99.5% later 2026	108	—	99.10%(med)	99.90%	97.00%	30µs	—	60ns	Tunable couplers, modular 12×9-qubit chiplets	link
IQM Garnet	20	—	99.51%(med)	99.92%	97.00%	30µs	—	30ns	Rotated square lattice (tunable)	link
Origin Wukong	72	—	98.50%	99.50%	—	15µs	2µs	30ns	Tunable-coupler 2D	Jan 6, 2024

Trapped Ion

System	Physical	Logical	2Q Fid.	1Q Fid.	SPAM	T1	T2	Gate	Connectivity	Source
Quantinuum Helios Industry record 2Q fidelity across all pairs	98	94	99.92%(med)	100.00%	99.80%	100.0s	1.0s	50µs	All-to-all (QCCD shuttling)	link
Quantinuum H2 QV 2^25 = 33,554,432 — industry record	56	12	99.91%(med)	100.00%	99.80%	—	1.0s	45µs	All-to-all (QCCD)	May 15, 2025
IonQ Forte Enterprise	36	—	99.60%(med)	99.98%	99.50%	1.0s	—	200µs	All-to-all (single chain)	Jun 1, 2025
IonQ Tempo Best fidelity on barium qubits via Oxford Ionics EQC tech	100	—	99.99%(best)	99.99%	99.96%	1.0s	—	100µs	All-to-all	link

Neutral Atom

System	Physical	Logical	2Q Fid.	1Q Fid.	SPAM	T1	T2	Gate	Connectivity	Source
Atom Computing Phoenix 1,225-site array; nuclear spin coherence ~40s	1,180	28	99.50%(med)	99.90%	99.00%	40.0s	—	1µs	Reconfigurable (movable tweezers)	Jan 20, 2025
QuEra Gemini	260	30	99.50%(med)	99.90%	99.00%	1.5s	—	1µs	Reconfigurable	Jun 20, 2025
Pasqal Orion Gamma 250-qubit target H1 2026	140	2	99.70%(med)	99.90%	99.00%	1.0s	—	1µs	Reconfigurable	Oct 1, 2025
Infleqtion Sqale Record CZ fidelity for long-lived NA qubits (excl. loss); 1,600-site lattice demoed	100	—	99.73%(best)	99.90%	99.00%	—	2.8s	1µs	Reconfigurable	Sep 1, 2025

Photonic

System	Physical	Logical	2Q Fid.	1Q Fid.	SPAM	T1	T2	Gate	Connectivity	Source
Xanadu Aurora 12 CV modes / 35 chips / 86B time-domain modes; first modular photonic system	—	12	—	—	—	—	—	—	Modular, networked (13 km fiber)	link
Quandela Belenos / MosaiQ 12	12	—	97.00%	99.00%	—	—	—	—	Photonic linear-optical	May 1, 2025

Silicon Spin

System	Physical	Logical	2Q Fid.	1Q Fid.	SPAM	T1	T2	Gate	Connectivity	Source
Diraq (imec foundry) First foundry-produced spin qubits to clear FT threshold; 99.9% SPAM	12	—	99.00%(med)	99.00%	99.90%	—	1ms	100ns	Nearest-neighbor (planar dots)	link
SQC (donor-based)	11	—	99.64%(best)	99.99%	99.00%	—	1.0s	1µs	Atom-precise STM lithography	Dec 15, 2025

Topological

System	Physical	Logical	2Q Fid.	1Q Fid.	SPAM	T1	T2	Gate	Connectivity	Source
Microsoft Majorana 1 Peer reviewers question existence of topological qubit; Nature editorial disclaimer	8	—	—	—	—	—	—	—	1D wire array (tetrons)	Feb 19, 2025

Annealing

System	Physical	Logical	2Q Fid.	1Q Fid.	SPAM	T1	T2	Gate	Connectivity	Source
D-Wave Advantage2 Analog annealer — not gate-based; 75% noise reduction vs Advantage1	4,400	—	—	—	—	—	—	—	Zephyr topology, degree 20 (40K+ couplers)	Jan 1, 2025

NV-Center

System	Physical	Logical	2Q Fid.	1Q Fid.	SPAM	T1	T2	Gate	Connectivity	Source
Quantum Brilliance QB-QDK2.0 ROOM TEMPERATURE — no cryogenics required	5	—	98.00%	99.00%	—	—	1ms	1µs	Nearest-neighbor on diamond	May 1, 2025

Rankings by dimension

Highest 2Q Gate Fidelity

1. IonQ Tempo99.99%
2. Quantinuum Helios99.92%
3. Quantinuum H299.91%
4. IBM Nighthawk99.90%
5. Google Willow99.88%
6. Infleqtion Sqale99.73%

Highest Physical Qubit Count

1. Atom Computing Phoenix1,180
2. IBM Condor1,121
3. QuEra Gemini260
4. IBM Heron R2 (Kingston/Aachen)156
5. Pasqal Orion Gamma140
6. IBM Nighthawk120

Most Logical Qubits

1. Quantinuum Helios94
2. QuEra Gemini30
3. Atom Computing Phoenix28
4. Quantinuum H212
5. Xanadu Aurora12
6. Pasqal Orion Gamma2

Longest Coherence (T2)

1. Infleqtion Sqale2.8s
2. Quantinuum Helios1.0s
3. Quantinuum H21.0s
4. SQC (donor-based)1.0s
5. Diraq (imec foundry)1ms
6. Quantum Brilliance QB-QDK2.01ms

Fastest Gates

1. Google Willow25ns
2. IQM Garnet30ns
3. Origin Wukong30ns
4. Rigetti Cepheus-1-108Q60ns
5. IBM Heron R2 (Kingston/Aachen)68ns
6. IBM Nighthawk70ns

Throughput (CLOPS)

1. IBM Heron R2 (Kingston/Aachen)330,000

Source citations link directly to peer-reviewed papers and primary vendor disclosures. See methodology for confidence-flag definitions and the contested-numbers explainer.