Data re-uploading QML: which simulator is fastest

Repeated encode-and-rotate classifier

8 simulators ran this circuit at 8 to 16 qubits, 21 workloads in all. TensorCircuit was the fastest in 15, Google qsim in 4 and Qulacs in 2. The tables below hold every time as it was measured, with the fastest in each row marked. Open this circuit in the interactive tool to see the hardware that could run it and what a run would cost.

Expectation value, global observable (all_z), depth 2

Time per run, by qubit count

QubitsPennyLanePennyLane LightningQiskitQiskit AerQiskit Aer (MPS)Google qsimQulacsTensorCircuit
8 qubits
PennyLane
10.4 ms
PennyLane Lightning
5.31 ms
Qiskit
3.65 ms
Qiskit Aer
2.33 ms
Qiskit Aer (MPS)
3.31 ms
Google qsim
3.99 ms
Qulacs
185 µs
TensorCircuit
388 µs
12 qubits
PennyLane
18.2 ms
PennyLane Lightning
7.58 ms
Qiskit
7.61 ms
Qiskit Aer
5.07 ms
Qiskit Aer (MPS)
6.67 ms
Google qsim
5.75 ms
Qulacs
2.06 ms
TensorCircuit
1.20 ms
16 qubits
PennyLane
81.9 ms
PennyLane Lightning
17.0 ms
Qiskit
72.6 ms
Qiskit Aer
38.8 ms
Qiskit Aer (MPS)
11.3 ms
Google qsim
9.07 ms
Qulacs
41.8 ms
TensorCircuit
43.8 ms

Expectation value, global observable (all_z), depth 2, batch of 32

Time per run, by qubit count

QubitsPennyLanePennyLane LightningQiskitQiskit AerQiskit Aer (MPS)Google qsimQulacsTensorCircuit
8 qubits
PennyLane
14.7 ms
PennyLane Lightning
53.2 ms
Qiskit
116 ms
Qiskit Aer
74.3 ms
Qiskit Aer (MPS)
106 ms
Google qsim
128 ms
Qulacs
6.97 ms
TensorCircuit
2.00 ms
12 qubits
PennyLane
121 ms
PennyLane Lightning
88.0 ms
Qiskit
246 ms
Qiskit Aer
162 ms
Qiskit Aer (MPS)
216 ms
Google qsim
186 ms
Qulacs
67.1 ms
TensorCircuit
59.8 ms
16 qubits
PennyLane
8.25 s
PennyLane Lightning
354 ms
Qiskit
2.26 s
Qiskit Aer
1.24 s
Qiskit Aer (MPS)
359 ms
Google qsim
289 ms
Qulacs
1.33 s
TensorCircuit
2.02 s

Expectation value, global observable (all_z), depth 2, batch of 256

Time per run, by qubit count

QubitsPennyLanePennyLane LightningQiskitQiskit AerQiskit Aer (MPS)Google qsimQulacsTensorCircuit
8 qubits
PennyLane
53.8 ms
PennyLane Lightning
398 ms
Qiskit
941 ms
Qiskit Aer
591 ms
Qiskit Aer (MPS)
839 ms
Google qsim
1.02 s
Qulacs
48.4 ms
TensorCircuit
14.7 ms
12 qubits
PennyLane
1.08 s
PennyLane Lightning
670 ms
Qiskit
1.94 s
Qiskit Aer
1.30 s
Qiskit Aer (MPS)
1.72 s
Google qsim
1.47 s
Qulacs
529 ms
TensorCircuit
497 ms
16 qubits
PennyLane
71.1 s
PennyLane Lightning
2.81 s
Qiskit
17.9 s
Qiskit Aer
9.89 s
Qiskit Aer (MPS)
2.87 s
Google qsim
2.35 s
Qulacs
10.6 s
TensorCircuit
16.9 s

Expectation value, global observable (all_z), depth 2, batch of 1024

Time per run, by qubit count

QubitsPennyLanePennyLane LightningQiskitQiskit AerQiskit Aer (MPS)Google qsimQulacsTensorCircuit
8 qubits
PennyLane
200 ms
PennyLane Lightning
1.80 s
Qiskit
3.74 s
Qiskit Aer
2.37 s
Qiskit Aer (MPS)
3.37 s
Google qsim
4.09 s
Qulacs
193 ms
TensorCircuit
107 ms
12 qubits
PennyLane
13.7 s
PennyLane Lightning
2.91 s
Qiskit
7.93 s
Qiskit Aer
5.23 s
Qiskit Aer (MPS)
6.96 s
Google qsim
5.89 s
Qulacs
2.13 s
TensorCircuit
2.17 s
16 qubits
PennyLane
skip
PennyLane Lightning
11.3 s
Qiskit
72.3 s
Qiskit Aer
39.6 s
Qiskit Aer (MPS)
11.4 s
Google qsim
9.24 s
Qulacs
42.7 s
TensorCircuit
error

Gradient, local observable, depth 2

Time per run, by qubit count

QubitsPennyLanePennyLane LightningQiskitQiskit AerQiskit Aer (MPS)Google qsimQulacsTensorCircuit
8 qubits
PennyLane
unsupported
PennyLane Lightning
unsupported
Qiskit
unsupported
Qiskit Aer
unsupported
Qiskit Aer (MPS)
unsupported
Google qsim
unsupported
Qulacs
606 µs
TensorCircuit
596 µs
12 qubits
PennyLane
unsupported
PennyLane Lightning
unsupported
Qiskit
unsupported
Qiskit Aer
unsupported
Qiskit Aer (MPS)
unsupported
Google qsim
unsupported
Qulacs
5.81 ms
TensorCircuit
1.16 ms
16 qubits
PennyLane
unsupported
PennyLane Lightning
unsupported
Qiskit
unsupported
Qiskit Aer
unsupported
Qiskit Aer (MPS)
unsupported
Google qsim
unsupported
Qulacs
117 ms
TensorCircuit
7.10 ms

Gradient, local observable, depth 2, batch of 32

Time per run, by qubit count

QubitsPennyLanePennyLane LightningQiskitQiskit AerQiskit Aer (MPS)Google qsimQulacsTensorCircuit
8 qubits
PennyLane
unsupported
PennyLane Lightning
unsupported
Qiskit
unsupported
Qiskit Aer
unsupported
Qiskit Aer (MPS)
unsupported
Google qsim
unsupported
Qulacs
19.3 ms
TensorCircuit
3.46 ms
12 qubits
PennyLane
unsupported
PennyLane Lightning
unsupported
Qiskit
unsupported
Qiskit Aer
unsupported
Qiskit Aer (MPS)
unsupported
Google qsim
unsupported
Qulacs
189 ms
TensorCircuit
19.9 ms
16 qubits
PennyLane
unsupported
PennyLane Lightning
unsupported
Qiskit
unsupported
Qiskit Aer
unsupported
Qiskit Aer (MPS)
unsupported
Google qsim
unsupported
Qulacs
3.76 s
TensorCircuit
424 ms

Gradient, local observable, depth 2, batch of 256

Time per run, by qubit count

QubitsPennyLanePennyLane LightningQiskitQiskit AerQiskit Aer (MPS)Google qsimQulacsTensorCircuit
8 qubits
PennyLane
unsupported
PennyLane Lightning
unsupported
Qiskit
unsupported
Qiskit Aer
unsupported
Qiskit Aer (MPS)
unsupported
Google qsim
unsupported
Qulacs
153 ms
TensorCircuit
26.0 ms
12 qubits
PennyLane
unsupported
PennyLane Lightning
unsupported
Qiskit
unsupported
Qiskit Aer
unsupported
Qiskit Aer (MPS)
unsupported
Google qsim
unsupported
Qulacs
1.47 s
TensorCircuit
243 ms
16 qubits
PennyLane
skip
PennyLane Lightning
unsupported
Qiskit
unsupported
Qiskit Aer
unsupported
Qiskit Aer (MPS)
unsupported
Google qsim
unsupported
Qulacs
30.0 s
TensorCircuit
3.51 s

Which quantum computers can run it

At depth 3, 1,000 shots

Quantum computerCan run it up toBorderline up toEstimated cost at that size
Can run it up to
16 qubits
Borderline up to
32 qubits
Estimated cost at that size
$200.30–$600.90
via AWS Braket
Can run it up to
32 qubits
Borderline up to
–
Estimated cost at that size
$200.30–$600.90
via AWS Braket
Can run it up to
8 qubits
Borderline up to
12 qubits
Estimated cost at that size
$23.80–$71.40
via AWS Braket
Can run it up to
96 qubits
Borderline up to
–
Estimated cost at that size
Not priced
this vendor publishes no rate card
Can run it up to
48 qubits
Borderline up to
–
Estimated cost at that size
Not priced
this vendor publishes no rate card
Can run it up to
48 qubits
Borderline up to
–
Estimated cost at that size
Not priced
this vendor publishes no rate card
Can run it up to
–
Borderline up to
–
Estimated cost at that size
–
Can run it up to
–
Borderline up to
–
Estimated cost at that size
–
Can run it up to
–
Borderline up to
–
Estimated cost at that size
–
Can run it up to
32 qubits
Borderline up to
96 qubits
Estimated cost at that size
from $0.015
via IBM Quantum. 0.01s at maximum throughput, so this is a floor
Can run it up to
48 qubits
Borderline up to
128 qubits
Estimated cost at that size
from $0.421
via IBM Quantum. 0.26s at maximum throughput, so this is a floor
Can run it up to
32 qubits
Borderline up to
96 qubits
Estimated cost at that size
from $0.421
via IBM Quantum. 0.26s at maximum throughput, so this is a floor
Can run it up to
4 qubits
Borderline up to
32 qubits
Estimated cost at that size
$0.725–$2.17
via AWS Braket
Can run it up to
24 qubits
Borderline up to
48 qubits
Estimated cost at that size
$1.90–$5.70
via AWS Braket
Can run it up to
16 qubits
Borderline up to
20 qubits
Estimated cost at that size
$1.75–$5.25
via AWS Braket

Sizes are the grid's own: 4, 8, 12, 16, 20, 24, 32, 48, 64, 96, 128, 156 qubits. A device can run a circuit when a clean run is more likely than not, and is borderline when it is better than one in fifty. The cost is for one run of 1,000 shots at the largest size in the row, as a band and not a quote.

The simulators

Each one across every circuit family

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.