Google qsim benchmark

Timed against 7 other simulators

Google qsim finished 138 of the 175 workloads it was run on and was the fastest of the 8 simulators in 44. At the median it took 4.0× the time of the fastest simulator in the same workload. What it is and how to install it is in its own documentation.

By circuit family

Where it is fast and where it is not

Circuit familyWorkloads finishedFastest inMedian time against the fastest
Workloads finished
12 of 16
Fastest in
0
Median time against the fastest
13.2×
Workloads finished
20 of 24
Fastest in
6
Median time against the fastest
5.5×
Workloads finished
23 of 27
Fastest in
9
Median time against the fastest
4.0×
Workloads finished
12 of 16
Fastest in
6
Median time against the fastest
2.6×
Workloads finished
20 of 24
Fastest in
10
Median time against the fastest
2.1×
Workloads finished
12 of 12
Fastest in
0
Median time against the fastest
3.3×
Workloads finished
15 of 19
Fastest in
5
Median time against the fastest
5.7×
Workloads finished
12 of 16
Fastest in
4
Median time against the fastest
4.7×
Workloads finished
12 of 21
Fastest in
4
Median time against the fastest
3.0×

Google qsim against each other simulator

On the workloads both finished

ComparisonWorkloads both finishedGoogle qsim faster inThe other faster inFaster at the median
Workloads both finished
119
Google qsim faster in
109
The other faster in
10
Faster at the median
Google qsim 2.7×
Workloads both finished
138
Google qsim faster in
107
The other faster in
31
Faster at the median
Google qsim 1.2×
Workloads both finished
121
Google qsim faster in
90
The other faster in
31
Faster at the median
Google qsim 1.4×
Workloads both finished
138
Google qsim faster in
73
The other faster in
65
Faster at the median
Google qsim 1.0×
Workloads both finished
120
Google qsim faster in
48
The other faster in
72
Faster at the median
Qiskit Aer (MPS) 1.1×
Workloads both finished
138
Google qsim faster in
64
The other faster in
74
Faster at the median
Qulacs 3.0×
Workloads both finished
118
Google qsim faster in
52
The other faster in
66
Faster at the median
TensorCircuit 2.9×

Where it produced no timing

The status the benchmark recorded

A simulator that did not finish a workload keeps the status the benchmark recorded for it, and is left out of that workload's ranking.

Recorded statusOperationWorkloads
unsupported
Operation
Gradient
Workloads
37

How this was measured

One host, one harness

Every simulator ran the same circuits on one host, single-threaded and core-pinned, driven through one common harness, so any overhead the harness adds is charged to the simulator. Each result was checked against a reference implementation: a simulator that ran fast but returned the wrong answer is left out of the ranking. Nothing is interpolated or rescaled. A time was measured, or it is not shown.

Run 2026-08-10, 8 cores, threads=1. numpy 2.4.6, scipy 1.18.0, torch 2.13.0, numba 0.66.0, pennylane 0.45.1, pennylane_lightning 0.45.0, qiskit 2.5.1, qulacs 0.6.14, tensorcircuit 1.8.0, cudaq 0.15.1. One host and one thread make the comparison reproducible. It is not a prediction for your laptop.

The interactive tool shows the same measurements for any workload you pick, with the quantum hardware that could run it and what a run would cost.

Hardware figures each carry the source they came from and the date they were measured. The oldest in this set is 626 days old. Costs are bands, not quotes.

Host: Linux-6.8.0-1051-aws-x86_64-with-glibc2.35, 8 cores, threads=1, core-pinned. numpy 2.4.6, scipy 1.18.0, torch 2.13.0, numba 0.66.0, pennylane 0.45.1, pennylane_lightning 0.45.0. Every simulator was driven through one common harness, so any harness binding overhead is charged to the simulator rather than hidden.

© 2026 EigenOrbit. All rights reserved.