Quantum Hardware Leaderboard
Compare quantum devices on one measurement at a time, from the public corpus . Each row shows that device's latest measurement, its date and its source. Vendor-reported rows are manufacturer claims, not independently reproduced here. For the long-run picture, see the tracker. How the corpus works →
Best 2Q fidelity
2Q gate fidelity
Median 2Q error
Median T1 (µs)
Median T2 (µs)
Median readout error
Algorithmic qubits (#AQ)
Quantum Volume
CLOPS
EPLG (error/layer)
Bell-state effective qubits
QFT accuracy
QAOA approx. ratio
Qubits
Best 2Q fidelity: Success rate of the chip's single best pair of qubits. The best pair flatters the chip; the typical one matters more.
Showing devices with a measurement in the last two years (48 older hidden).
Include older devices
| Rank | Maker | Device | Best 2Q fidelity | Source / licence | Measured |
|---|---|---|---|---|---|
| #1 | ibm | ibm_berlin | 0.9995 fails about 1 in 1,920 | IBM · Apache-2.0 | 2026-04-17 |
| #2 | ibm | ibm_boston | 0.9993 fails about 1 in 1,530 | IBM · Apache-2.0 | 2026-04-17 |
| #3 | ibm | ibm_aachen | 0.9993 fails about 1 in 1,450 | IBM · Apache-2.0 | 2026-04-17 |
| #4 | ibm | ibm_pittsburgh | 0.9992 fails about 1 in 1,270 | IBM · Apache-2.0 | 2026-04-17 |
| #5 | ibm | ibm_phoenix | 0.9992 fails about 1 in 1,220 | IBM · Apache-2.0 | 2026-11-10 |
| #6 | ibm | ibm_chicago | 0.9992 fails about 1 in 1,200 | IBM · Apache-2.0 | 2026-11-10 |
| #7 | ibm | ibm_kingston | 0.9991 fails about 1 in 1,150 | IBM · Apache-2.0 | 2026-04-15 |
| #8 | ibm | ibm_marrakesh | 0.999 fails about 1 in 970 | IBM · Apache-2.0 | 2025-02-26 |
| #9 | ibm | ibm_miami | 0.9989 fails about 1 in 950 | IBM · Apache-2.0 | 2026-04-17 |
| #10 | ibm | ibm_torino | 0.9985 fails about 1 in 670 | IBM · Apache-2.0 | 2025-02-26 |
| #11 | ibm | ibm_fez | 0.9983 fails about 1 in 580 | IBM · Apache-2.0 | 2025-02-26 |
| #12 | ibm | ibm_brussels | 0.9969 fails about 1 in 320 | IBM · Apache-2.0 | 2026-04-17 |
| #13 | ibm | ibm_kyiv | 0.9969 fails about 1 in 320 | IBM · Apache-2.0 | 2025-02-26 |
| #14 | ibm | ibm_quebec | 0.9969 fails about 1 in 320 | IBM · Apache-2.0 | 2025-02-26 |
| #15 | ibm | ibm_strasbourg | 0.9968 fails about 1 in 310 | IBM · Apache-2.0 | 2026-04-17 |
| #16 | ibm | ibm_sherbrooke | 0.9965 fails about 1 in 290 | IBM · Apache-2.0 | 2025-02-26 |
| #17 | ibm | ibm_brisbane | 0.9963 fails about 1 in 270 | IBM · Apache-2.0 | 2025-02-26 |
| #18 | ibm | ibm_hanoi | 0.9962 fails about 1 in 260 | IBM · Apache-2.0 | 2025-02-26 |
| #19 | ibm | ibm_kawasaki | 0.9959 fails about 1 in 240 | IBM · Apache-2.0 | 2025-02-26 |
| #20 | ibm | ibm_cusco | 0.9945 fails about 1 in 180 | IBM · Apache-2.0 | 2025-02-26 |
Sources & attribution.
IBM rows are real device calibration snapshots published in Qiskit's fake backends (Apache-2.0), each with its true date.
Metriq rows are community benchmark runs from metriq.info (CC-BY-4.0).
Vendor-reported rows are specifications published by the manufacturer and are shown as claims, with the source link recorded in each snapshot's provenance.
Metrics differ in what they measure and how they're obtained, so compare within a metric and mind the source badge.