IBM ibm_marrakesh

IBM · Apache-2.0 Metriq · CC-BY-4.0 last capture 2026-02-17

ibm_marrakesh is a 156-qubit IBM quantum computer, with records from 2025 to 2026.

On its latest calibration (26 February 2025), a typical two-qubit operation failed about 1 in 320, better than 87% of the 67 other devices in this record.

Every figure below is the device's latest measurement of that metric, with its own date and source. See how the whole field is progressing →

Best 2Q fidelity
IBM · Apache-2.0
0.998964
fails about 1 in 970

Success rate of the chip's single best pair of qubits. The best pair flatters the chip; the typical one matters more.

Better than 90% of 67 other devices with this measurement
Measured 2025-02-26
Median 2Q error
IBM · Apache-2.0
0.00317177
fails about 1 in 320

How often a typical two-qubit operation on this chip fails, from the vendor's own calibration.

Better than 87% of 67 other devices with this measurement
Measured 2025-02-26
Median T1 (µs)
IBM · Apache-2.0
197.36

How long a typical qubit holds its state before it decays, in microseconds. Longer leaves more time to compute.

Better than 77% of 69 other devices with this measurement
Measured 2025-02-26
Median T2 (µs)
IBM · Apache-2.0
118.43

How long a typical qubit keeps its delicate quantum phase, in microseconds.

Better than 70% of 69 other devices with this measurement
Measured 2025-02-26
Median readout error
IBM · Apache-2.0
0.00952148
fails about 1 in 110

How often reading a typical qubit's value at the end gives the wrong answer.

Better than 88% of 69 other devices with this measurement
Measured 2025-02-26
CLOPS
Metriq · CC-BY-4.0
353642

Circuit layers per second: how fast the machine gets through work, overheads included.

Better than 20% of 5 other devices with this measurement
Measured 2026-02-17
QFT accuracy
Metriq · CC-BY-4.0
0

How well the machine runs a small quantum Fourier transform benchmark: 1 is perfect, 0 is no better than noise.

Better than 0% of 10 other devices with this measurement
Measured 2026-02-09
QAOA approx. ratio
Metriq · CC-BY-4.0
0.573133

How much better than random guessing a simple optimisation algorithm does on this machine: 1 is perfect, 0 is no better than chance.

Better than 70% of 10 other devices with this measurement
Measured 2026-02-17
Qubits
Metriq · CC-BY-4.0
156

Physical qubits on the device. More only helps if they are good enough to use together.

Measured 2026-02-17

History

Each point is one dated measurement (4 snapshots in total).

Best 2Q fidelity
Median 2Q error
Median T1 (µs)
Median T2 (µs)
Median readout error
Qubits

Provenance

Source
Metriq · CC-BY-4.0
Licence
CC-BY-4.0 · metriq.info
Raw redistributable
yes
Captured
2026-02-17T11:59:22

Cite this page

@misc{provenova_ibm_ibm_marrakesh,
  title        = {IBM ibm_marrakesh — calibration & benchmark history},
  howpublished = {\url{https://provenova.net/hardware/ibm/ibm_marrakesh}},
  publisher    = {Provenova},
  note         = {accessed 2026-11-19}
}

Compare with

vs iqm iqm_emerald 2 shared metrics vs origin wukong_72 2 shared metrics vs quantinuum h2-2 2 shared metrics vs rigetti cepheus-1-108q 2 shared metrics vs ibm ibm_aachen 5 shared metrics vs ibm ibm_algiers 5 shared metrics vs ibm ibm_auckland 5 shared metrics vs ibm ibm_berlin 5 shared metrics

Data: Metriq community benchmarks (metriq.info, CC-BY-4.0). Rankings on the live leaderboard.