Quantum computing records: error rates & qubit counts

The full record history for every tracked hardware platform, every data point with its source.

As of March 2026, the lowest demonstrated two-qubit gate error rate is 0.0003 (trapped-ion, July 2024) and the largest disclosed single-device physical qubit count is 1,180 (neutral atom, October 2023). The two records sit on different platforms, which is the current shape of the field: no single hardware type leads on both fidelity and scale. Records are single-device results from the linked sources, not fleet averages, and a record on one metric says nothing about the same machine's standing on the other.

Current records at a glance

Platform	Record error rate	Record qubits	Error halves every	Qubits double every
Neutral Atom	0.00136	1,180	~1.5 y	~1.5 y
Superconducting	0.000519	1,121	~1.8 y	~1.4 y
Trapped-Ion	0.0003	98	~5.8 y	~2.3 y

Halving and doubling periods are log-linear fits of few-point record series; they describe the record history and are not predictions. The fastest error-rate improvement on record is neutral atom (halving every ~1.5 years); the fastest qubit-count growth is superconducting (doubling every ~1.4 years).

Neutral Atom record history

As of March 2026, the record two-qubit gate error rate on neutral atom hardware is 0.00136 (March 2026) and the record physical qubit count is 1,180 (October 2023). On a log-linear fit of the 4-point record series, the error rate halves every ~1.5 years and the qubit count doubles every ~1.5 years.

Two-qubit gate error rate

Date	Error rate	Source
August 2019	0.026	arxiv.org/abs/1908.06101
April 2023	0.005	arxiv.org/abs/2304.05420
July 2024	0.0029	arxiv.org/abs/2407.20184
March 2026	0.00136	arxiv.org/abs/2603.15561

Physical qubit count

Date	Qubits	Source
July 2017	51	arxiv.org/abs/1707.04344
July 2021	100	arxiv.org/abs/2108.04790
October 2023	1,180	www.hpcwire.com/2023/10/24/atom-computing-wins-the-race-to-1000-qubits/

Superconducting record history

As of September 2025, the record two-qubit gate error rate on superconducting hardware is 0.000519 (September 2025) and the record physical qubit count is 1,121 (December 2023). On a log-linear fit of the 3-point record series, the error rate halves every ~1.8 years and the qubit count doubles every ~1.4 years.

Two-qubit gate error rate

Date	Error rate	Source
October 2012	0.0653	arxiv.org/abs/1210.7011
July 2019	0.0034	arxiv.org/abs/1907.02510
September 2025	0.000519	www.linkedin.com/posts/jay-gambetta-a274753a_ibmquantum-quantumcomputing-activity-7330779528195100672-1TCI/

Physical qubit count

Date	Qubits	Source
February 2014	5	arxiv.org/abs/1402.4848
November 2014	9	arxiv.org/abs/1411.7403
May 2017	17	uk.newsroom.ibm.com/2017-05-17-IBM-Builds-Its-Most-Powerful-Universal-Quantum-Computing-Processors
March 2018	72	research.google/blog/a-preview-of-bristlecone-googles-new-quantum-processor/
November 2021	127	www.ibm.com/quantum/blog/127-qubit-quantum-processor-eagle
November 2022	433	newsroom.ibm.com/2022-11-09-IBM-Unveils-400-Qubit-Plus-Quantum-Processor-and-Next-Generation-IBM-Quantum-System-Two
December 2023	1,121	www.ibm.com/quantum/blog/quantum-roadmap-2033

Trapped-Ion record history

As of November 2025, the record two-qubit gate error rate on trapped-ion hardware is 0.0003 (July 2024) and the record physical qubit count is 98 (November 2025). On a log-linear fit of the 3-point record series, the error rate halves every ~5.8 years and the qubit count doubles every ~2.3 years.

Two-qubit gate error rate

Date	Error rate	Source
April 2016	0.00084	arxiv.org/abs/1604.00032
May 2021	0.0006	arxiv.org/abs/2105.05828
July 2024	0.0003	arxiv.org/abs/2407.07694

Physical qubit count

Date	Qubits	Source
March 2016	5	arxiv.org/abs/1603.04512
February 2022	24	www.aqt.eu/about/
May 2023	32	arxiv.org/abs/2305.03766
June 2024	56	arxiv.org/abs/2406.02501
November 2025	98	arxiv.org/abs/2511.05465