PennyLane vs Qulacs
The same circuits, one host
PennyLane and Qulacs both finished 147 of the 175 workloads measured here. PennyLane was faster in 0 of them and Qulacs in 147. At the median Qulacs was 10.6× faster.
By circuit family
Who was faster where
| Circuit family | Workloads both finished | PennyLane faster in | Qulacs faster in | Faster at the median |
|---|---|---|---|---|
| Workloads both finished 14 | PennyLane faster in 0 | Qulacs faster in 14 | Faster at the median Qulacs 12.5× | |
| Workloads both finished 22 | PennyLane faster in 0 | Qulacs faster in 22 | Faster at the median Qulacs 18.1× | |
| Workloads both finished 25 | PennyLane faster in 0 | Qulacs faster in 25 | Faster at the median Qulacs 10.4× | |
| Workloads both finished 14 | PennyLane faster in 0 | Qulacs faster in 14 | Faster at the median Qulacs 19.8× | |
| Workloads both finished 22 | PennyLane faster in 0 | Qulacs faster in 22 | Faster at the median Qulacs 10.9× | |
| Workloads both finished 8 | PennyLane faster in 0 | Qulacs faster in 8 | Faster at the median Qulacs 3.0× | |
| Workloads both finished 17 | PennyLane faster in 0 | Qulacs faster in 17 | Faster at the median Qulacs 12.1× | |
| Workloads both finished 14 | PennyLane faster in 0 | Qulacs faster in 14 | Faster at the median Qulacs 11.8× | |
| Workloads both finished 11 | PennyLane faster in 0 | Qulacs faster in 11 | Faster at the median Qulacs 2.1× |
Local circuit dynamics
Brickwork nearest-neighbour layers: lattice models, quench dynamics
| Workload | Qubits | PennyLane | Qulacs | Faster |
|---|---|---|---|---|
Expectation value, local observable, depth 3 | Qubits 4 | PennyLane 2.07 ms | Qulacs 32.1 µs | Faster Qulacs 64.5× |
Expectation value, local observable, depth 3 | Qubits 8 | PennyLane 3.73 ms | Qulacs 73.1 µs | Faster Qulacs 51.1× |
Expectation value, local observable, depth 3 | Qubits 12 | PennyLane 6.19 ms | Qulacs 585 µs | Faster Qulacs 10.6× |
Expectation value, local observable, depth 3 | Qubits 16 | PennyLane 30.3 ms | Qulacs 10.7 ms | Faster Qulacs 2.8× |
Expectation value, local observable, depth 3 | Qubits 20 | PennyLane 697 ms | Qulacs 206 ms | Faster Qulacs 3.4× |
Expectation value, local observable, depth 3 | Qubits 24 | PennyLane skip | Qulacs 4.76 s | Faster – |
Expectation value, global observable (all_z), depth 3 | Qubits 4 | PennyLane 2.48 ms | Qulacs 42.5 µs | Faster Qulacs 58.3× |
Expectation value, global observable (all_z), depth 3 | Qubits 8 | PennyLane 4.76 ms | Qulacs 119 µs | Faster Qulacs 39.9× |
Expectation value, global observable (all_z), depth 3 | Qubits 12 | PennyLane 8.07 ms | Qulacs 1.38 ms | Faster Qulacs 5.9× |
Expectation value, global observable (all_z), depth 3 | Qubits 16 | PennyLane 43.5 ms | Qulacs 26.2 ms | Faster Qulacs 1.7× |
Expectation value, global observable (all_z), depth 3 | Qubits 20 | PennyLane 891 ms | Qulacs 535 ms | Faster Qulacs 1.7× |
Expectation value, global observable (all_z), depth 3 | Qubits 24 | PennyLane skip | Qulacs 10.7 s | Faster – |
Gradient, local observable, depth 3 | Qubits 4 | PennyLane 10.7 ms | Qulacs 117 µs | Faster Qulacs 91.7× |
Gradient, local observable, depth 3 | Qubits 8 | PennyLane 22.3 ms | Qulacs 307 µs | Faster Qulacs 72.6× |
Gradient, local observable, depth 3 | Qubits 12 | PennyLane 43.4 ms | Qulacs 3.00 ms | Faster Qulacs 14.5× |
Gradient, local observable, depth 3 | Qubits 16 | PennyLane 268 ms | Qulacs 59.6 ms | Faster Qulacs 4.5× |
Hamiltonian time evolution
Trotterised spin chain: ZZ + XX couplings with local fields
| Workload | Qubits | PennyLane | Qulacs | Faster |
|---|---|---|---|---|
Expectation value, local observable, depth 3 | Qubits 4 | PennyLane 9.46 ms | Qulacs 104 µs | Faster Qulacs 90.8× |
Expectation value, local observable, depth 3 | Qubits 8 | PennyLane 23.2 ms | Qulacs 276 µs | Faster Qulacs 84.3× |
Expectation value, local observable, depth 3 | Qubits 12 | PennyLane 43.8 ms | Qulacs 1.85 ms | Faster Qulacs 23.7× |
Expectation value, local observable, depth 3 | Qubits 16 | PennyLane 208 ms | Qulacs 34.0 ms | Faster Qulacs 6.1× |
Expectation value, local observable, depth 3 | Qubits 20 | PennyLane 4.57 s | Qulacs 711 ms | Faster Qulacs 6.4× |
Expectation value, local observable, depth 3 | Qubits 24 | PennyLane skip | Qulacs 21.1 s | Faster – |
Expectation value, global observable (all_z), depth 3 | Qubits 4 | PennyLane 9.82 ms | Qulacs 113 µs | Faster Qulacs 87.2× |
Expectation value, global observable (all_z), depth 3 | Qubits 8 | PennyLane 24.4 ms | Qulacs 321 µs | Faster Qulacs 76.1× |
Expectation value, global observable (all_z), depth 3 | Qubits 12 | PennyLane 46.3 ms | Qulacs 2.51 ms | Faster Qulacs 18.5× |
Expectation value, global observable (all_z), depth 3 | Qubits 16 | PennyLane 223 ms | Qulacs 48.2 ms | Faster Qulacs 4.6× |
Expectation value, global observable (all_z), depth 3 | Qubits 20 | PennyLane 4.86 s | Qulacs 1.02 s | Faster Qulacs 4.8× |
Expectation value, global observable (all_z), depth 3 | Qubits 24 | PennyLane skip | Qulacs 26.9 s | Faster – |
Expectation value, global observable (pauli_mix), depth 3 | Qubits 8 | PennyLane 25.0 ms | Qulacs 349 µs | Faster Qulacs 71.8× |
Expectation value, global observable (pauli_mix), depth 3 | Qubits 12 | PennyLane 46.9 ms | Qulacs 3.17 ms | Faster Qulacs 14.8× |
Expectation value, global observable (pauli_mix), depth 3 | Qubits 16 | PennyLane 236 ms | Qulacs 61.5 ms | Faster Qulacs 3.8× |
Expectation value, global observable (pauli_mix), depth 3 | Qubits 20 | PennyLane 5.11 s | Qulacs 1.28 s | Faster Qulacs 4.0× |
Expectation value, global observable (zz_sum), depth 3 | Qubits 8 | PennyLane 23.7 ms | Qulacs 305 µs | Faster Qulacs 77.7× |
Expectation value, global observable (zz_sum), depth 3 | Qubits 12 | PennyLane 44.4 ms | Qulacs 2.50 ms | Faster Qulacs 17.7× |
Expectation value, global observable (zz_sum), depth 3 | Qubits 16 | PennyLane 211 ms | Qulacs 48.1 ms | Faster Qulacs 4.4× |
Expectation value, global observable (zz_sum), depth 3 | Qubits 20 | PennyLane 4.76 s | Qulacs 1.01 s | Faster Qulacs 4.7× |
Gradient, local observable, depth 3 | Qubits 4 | PennyLane 29.5 ms | Qulacs 424 µs | Faster Qulacs 69.7× |
Gradient, local observable, depth 3 | Qubits 8 | PennyLane 72.7 ms | Qulacs 1.17 ms | Faster Qulacs 62.0× |
Gradient, local observable, depth 3 | Qubits 12 | PennyLane 212 ms | Qulacs 9.19 ms | Faster Qulacs 23.1× |
Gradient, local observable, depth 3 | Qubits 16 | PennyLane 1.64 s | Qulacs 181 ms | Faster Qulacs 9.1× |
MaxCut on a ring
QAOA over a cycle graph
| Workload | Qubits | PennyLane | Qulacs | Faster |
|---|---|---|---|---|
Expectation value, local observable, depth 3 | Qubits 4 | PennyLane 5.17 ms | Qulacs 63.1 µs | Faster Qulacs 81.9× |
Expectation value, local observable, depth 3 | Qubits 8 | PennyLane 10.6 ms | Qulacs 154 µs | Faster Qulacs 68.8× |
Expectation value, local observable, depth 3 | Qubits 12 | PennyLane 17.8 ms | Qulacs 1.14 ms | Faster Qulacs 15.6× |
Expectation value, local observable, depth 3 | Qubits 16 | PennyLane 82.9 ms | Qulacs 20.9 ms | Faster Qulacs 4.0× |
Expectation value, local observable, depth 3 | Qubits 20 | PennyLane 1.82 s | Qulacs 433 ms | Faster Qulacs 4.2× |
Expectation value, local observable, depth 3 | Qubits 24 | PennyLane skip | Qulacs 10.6 s | Faster – |
Expectation value, global observable (all_z), depth 3 | Qubits 4 | PennyLane 5.56 ms | Qulacs 72.0 µs | Faster Qulacs 77.2× |
Expectation value, global observable (all_z), depth 3 | Qubits 8 | PennyLane 11.3 ms | Qulacs 197 µs | Faster Qulacs 57.2× |
Expectation value, global observable (all_z), depth 3 | Qubits 12 | PennyLane 19.9 ms | Qulacs 1.91 ms | Faster Qulacs 10.4× |
Expectation value, global observable (all_z), depth 3 | Qubits 16 | PennyLane 95.7 ms | Qulacs 36.9 ms | Faster Qulacs 2.6× |
Expectation value, global observable (all_z), depth 3 | Qubits 20 | PennyLane 2.05 s | Qulacs 740 ms | Faster Qulacs 2.8× |
Expectation value, global observable (all_z), depth 3 | Qubits 24 | PennyLane skip | Qulacs 16.6 s | Faster – |
Expectation value, global observable (pauli_mix), depth 3 | Qubits 8 | PennyLane 12.7 ms | Qulacs 234 µs | Faster Qulacs 54.3× |
Expectation value, global observable (pauli_mix), depth 3 | Qubits 12 | PennyLane 22.7 ms | Qulacs 2.61 ms | Faster Qulacs 8.7× |
Expectation value, global observable (pauli_mix), depth 3 | Qubits 16 | PennyLane 116 ms | Qulacs 52.1 ms | Faster Qulacs 2.2× |
Expectation value, global observable (pauli_mix), depth 3 | Qubits 20 | PennyLane 2.39 s | Qulacs 1.03 s | Faster Qulacs 2.3× |
Expectation value, global observable (zz_sum), depth 3 | Qubits 8 | PennyLane 11.7 ms | Qulacs 191 µs | Faster Qulacs 61.5× |
Expectation value, global observable (zz_sum), depth 3 | Qubits 12 | PennyLane 20.2 ms | Qulacs 1.95 ms | Faster Qulacs 10.4× |
Expectation value, global observable (zz_sum), depth 3 | Qubits 16 | PennyLane 93.4 ms | Qulacs 38.6 ms | Faster Qulacs 2.4× |
Expectation value, global observable (zz_sum), depth 3 | Qubits 20 | PennyLane 1.88 s | Qulacs 762 ms | Faster Qulacs 2.5× |
Gradient, local observable, depth 3 | Qubits 4 | PennyLane 23.6 ms | Qulacs 255 µs | Faster Qulacs 92.4× |
Gradient, local observable, depth 3 | Qubits 8 | PennyLane 48.6 ms | Qulacs 681 µs | Faster Qulacs 71.4× |
Gradient, local observable, depth 3 | Qubits 12 | PennyLane 113 ms | Qulacs 6.05 ms | Faster Qulacs 18.8× |
Gradient, local observable, depth 3 | Qubits 16 | PennyLane 764 ms | Qulacs 120 ms | Faster Qulacs 6.4× |
Statevector, global observable (all_z), depth 3 | Qubits 8 | PennyLane 10.9 ms | Qulacs 155 µs | Faster Qulacs 70.1× |
Statevector, global observable (all_z), depth 3 | Qubits 12 | PennyLane 19.1 ms | Qulacs 1.13 ms | Faster Qulacs 16.9× |
Statevector, global observable (all_z), depth 3 | Qubits 16 | PennyLane 88.1 ms | Qulacs 21.0 ms | Faster Qulacs 4.2× |
MaxCut, arbitrary graph
QAOA over a random-edge graph
| Workload | Qubits | PennyLane | Qulacs | Faster |
|---|---|---|---|---|
Expectation value, local observable, depth 3 | Qubits 4 | PennyLane 6.19 ms | Qulacs 71.4 µs | Faster Qulacs 86.7× |
Expectation value, local observable, depth 3 | Qubits 8 | PennyLane 14.4 ms | Qulacs 188 µs | Faster Qulacs 76.8× |
Expectation value, local observable, depth 3 | Qubits 12 | PennyLane 28.0 ms | Qulacs 1.51 ms | Faster Qulacs 18.6× |
Expectation value, local observable, depth 3 | Qubits 16 | PennyLane 115 ms | Qulacs 25.8 ms | Faster Qulacs 4.4× |
Expectation value, local observable, depth 3 | Qubits 20 | PennyLane 2.79 s | Qulacs 588 ms | Faster Qulacs 4.7× |
Expectation value, local observable, depth 3 | Qubits 24 | PennyLane skip | Qulacs 16.1 s | Faster – |
Expectation value, global observable (all_z), depth 3 | Qubits 4 | PennyLane 6.57 ms | Qulacs 81.0 µs | Faster Qulacs 81.1× |
Expectation value, global observable (all_z), depth 3 | Qubits 8 | PennyLane 15.4 ms | Qulacs 239 µs | Faster Qulacs 64.5× |
Expectation value, global observable (all_z), depth 3 | Qubits 12 | PennyLane 30.3 ms | Qulacs 2.39 ms | Faster Qulacs 12.7× |
Expectation value, global observable (all_z), depth 3 | Qubits 16 | PennyLane 126 ms | Qulacs 46.3 ms | Faster Qulacs 2.7× |
Expectation value, global observable (all_z), depth 3 | Qubits 20 | PennyLane 3.07 s | Qulacs 1.00 s | Faster Qulacs 3.1× |
Expectation value, global observable (all_z), depth 3 | Qubits 24 | PennyLane skip | Qulacs 24.2 s | Faster – |
Gradient, local observable, depth 3 | Qubits 4 | PennyLane 26.9 ms | Qulacs 287 µs | Faster Qulacs 93.5× |
Gradient, local observable, depth 3 | Qubits 8 | PennyLane 63.3 ms | Qulacs 873 µs | Faster Qulacs 72.5× |
Gradient, local observable, depth 3 | Qubits 12 | PennyLane 170 ms | Qulacs 8.07 ms | Faster Qulacs 21.1× |
Gradient, local observable, depth 3 | Qubits 16 | PennyLane 1.08 s | Qulacs 148 ms | Faster Qulacs 7.3× |
Sampling / IQP
Diagonal phase circuit between Hadamard layers
| Workload | Qubits | PennyLane | Qulacs | Faster |
|---|---|---|---|---|
Expectation value, local observable, depth 3 | Qubits 4 | PennyLane 5.68 ms | Qulacs 68.8 µs | Faster Qulacs 82.6× |
Expectation value, local observable, depth 3 | Qubits 8 | PennyLane 11.3 ms | Qulacs 156 µs | Faster Qulacs 72.7× |
Expectation value, local observable, depth 3 | Qubits 12 | PennyLane 19.7 ms | Qulacs 1.04 ms | Faster Qulacs 18.9× |
Expectation value, local observable, depth 3 | Qubits 16 | PennyLane 95.5 ms | Qulacs 18.5 ms | Faster Qulacs 5.2× |
Expectation value, local observable, depth 3 | Qubits 20 | PennyLane 2.15 s | Qulacs 392 ms | Faster Qulacs 5.5× |
Expectation value, local observable, depth 3 | Qubits 24 | PennyLane skip | Qulacs 10.7 s | Faster – |
Expectation value, global observable (all_z), depth 3 | Qubits 4 | PennyLane 6.14 ms | Qulacs 78.2 µs | Faster Qulacs 78.5× |
Expectation value, global observable (all_z), depth 3 | Qubits 8 | PennyLane 12.6 ms | Qulacs 203 µs | Faster Qulacs 62.2× |
Expectation value, global observable (all_z), depth 3 | Qubits 12 | PennyLane 21.6 ms | Qulacs 1.97 ms | Faster Qulacs 11.0× |
Expectation value, global observable (all_z), depth 3 | Qubits 16 | PennyLane 105 ms | Qulacs 38.7 ms | Faster Qulacs 2.7× |
Expectation value, global observable (all_z), depth 3 | Qubits 20 | PennyLane 2.48 s | Qulacs 797 ms | Faster Qulacs 3.1× |
Expectation value, global observable (all_z), depth 3 | Qubits 24 | PennyLane skip | Qulacs 19.1 s | Faster – |
Expectation value, global observable (pauli_mix), depth 3 | Qubits 8 | PennyLane 13.7 ms | Qulacs 239 µs | Faster Qulacs 57.1× |
Expectation value, global observable (pauli_mix), depth 3 | Qubits 12 | PennyLane 24.5 ms | Qulacs 2.67 ms | Faster Qulacs 9.2× |
Expectation value, global observable (pauli_mix), depth 3 | Qubits 16 | PennyLane 127 ms | Qulacs 53.4 ms | Faster Qulacs 2.4× |
Expectation value, global observable (pauli_mix), depth 3 | Qubits 20 | PennyLane 2.74 s | Qulacs 1.09 s | Faster Qulacs 2.5× |
Expectation value, global observable (zz_sum), depth 3 | Qubits 8 | PennyLane 12.8 ms | Qulacs 196 µs | Faster Qulacs 65.1× |
Expectation value, global observable (zz_sum), depth 3 | Qubits 12 | PennyLane 22.2 ms | Qulacs 2.04 ms | Faster Qulacs 10.9× |
Expectation value, global observable (zz_sum), depth 3 | Qubits 16 | PennyLane 108 ms | Qulacs 39.9 ms | Faster Qulacs 2.7× |
Expectation value, global observable (zz_sum), depth 3 | Qubits 20 | PennyLane 2.21 s | Qulacs 856 ms | Faster Qulacs 2.6× |
Gradient, local observable, depth 3 | Qubits 4 | PennyLane 20.1 ms | Qulacs 272 µs | Faster Qulacs 73.9× |
Gradient, local observable, depth 3 | Qubits 8 | PennyLane 43.1 ms | Qulacs 701 µs | Faster Qulacs 61.5× |
Gradient, local observable, depth 3 | Qubits 12 | PennyLane 112 ms | Qulacs 5.38 ms | Faster Qulacs 20.9× |
Gradient, local observable, depth 3 | Qubits 16 | PennyLane 742 ms | Qulacs 105 ms | Faster Qulacs 7.1× |
Quantum kernel feature map
Data-encoding circuit for kernel methods
| Workload | Qubits | PennyLane | Qulacs | Faster |
|---|---|---|---|---|
Expectation value, global observable (all_z), depth 2 | Qubits 8 | PennyLane 8.00 ms | Qulacs 107 µs | Faster Qulacs 75.1× |
Expectation value, global observable (all_z), depth 2 | Qubits 12 | PennyLane 12.7 ms | Qulacs 895 µs | Faster Qulacs 14.2× |
Expectation value, global observable (all_z), depth 2 | Qubits 16 | PennyLane skip | Qulacs 16.9 ms | Faster – |
Expectation value, global observable (all_z), depth 2, batch of 32 | Qubits 8 | PennyLane 10.6 ms | Qulacs 3.36 ms | Faster Qulacs 3.2× |
Expectation value, global observable (all_z), depth 2, batch of 32 | Qubits 12 | PennyLane 76.1 ms | Qulacs 28.3 ms | Faster Qulacs 2.7× |
Expectation value, global observable (all_z), depth 2, batch of 32 | Qubits 16 | PennyLane skip | Qulacs 539 ms | Faster – |
Expectation value, global observable (all_z), depth 2, batch of 256 | Qubits 8 | PennyLane 30.3 ms | Qulacs 26.9 ms | Faster Qulacs 1.1× |
Expectation value, global observable (all_z), depth 2, batch of 256 | Qubits 12 | PennyLane 652 ms | Qulacs 226 ms | Faster Qulacs 2.9× |
Expectation value, global observable (all_z), depth 2, batch of 256 | Qubits 16 | PennyLane skip | Qulacs 4.31 s | Faster – |
Expectation value, global observable (all_z), depth 2, batch of 1024 | Qubits 8 | PennyLane 110 ms | Qulacs 109 ms | Faster Qulacs 1.0× |
Expectation value, global observable (all_z), depth 2, batch of 1024 | Qubits 12 | PennyLane 7.92 s | Qulacs 915 ms | Faster Qulacs 8.7× |
Expectation value, global observable (all_z), depth 2, batch of 1024 | Qubits 16 | PennyLane skip | Qulacs 17.3 s | Faster – |
Hardware-efficient VQE
RY/RZ layers with a CNOT ring
| Workload | Qubits | PennyLane | Qulacs | Faster |
|---|---|---|---|---|
Expectation value, local observable, depth 3 | Qubits 4 | PennyLane 3.43 ms | Qulacs 52.5 µs | Faster Qulacs 65.3× |
Expectation value, local observable, depth 3 | Qubits 8 | PennyLane 6.83 ms | Qulacs 126 µs | Faster Qulacs 54.1× |
Expectation value, local observable, depth 3 | Qubits 12 | PennyLane 12.0 ms | Qulacs 1.03 ms | Faster Qulacs 11.6× |
Expectation value, local observable, depth 3 | Qubits 16 | PennyLane 61.6 ms | Qulacs 19.2 ms | Faster Qulacs 3.2× |
Expectation value, local observable, depth 3 | Qubits 20 | PennyLane 1.41 s | Qulacs 386 ms | Faster Qulacs 3.7× |
Expectation value, local observable, depth 3 | Qubits 24 | PennyLane skip | Qulacs 8.90 s | Faster – |
Expectation value, global observable (all_z), depth 3 | Qubits 4 | PennyLane 3.91 ms | Qulacs 63.7 µs | Faster Qulacs 61.3× |
Expectation value, global observable (all_z), depth 3 | Qubits 8 | PennyLane 7.93 ms | Qulacs 172 µs | Faster Qulacs 46.2× |
Expectation value, global observable (all_z), depth 3 | Qubits 12 | PennyLane 14.0 ms | Qulacs 1.97 ms | Faster Qulacs 7.1× |
Expectation value, global observable (all_z), depth 3 | Qubits 16 | PennyLane 72.0 ms | Qulacs 40.2 ms | Faster Qulacs 1.8× |
Expectation value, global observable (all_z), depth 3 | Qubits 20 | PennyLane 1.54 s | Qulacs 808 ms | Faster Qulacs 1.9× |
Expectation value, global observable (all_z), depth 3 | Qubits 24 | PennyLane skip | Qulacs 17.1 s | Faster – |
Gradient, local observable, depth 3 | Qubits 4 | PennyLane 19.4 ms | Qulacs 206 µs | Faster Qulacs 93.9× |
Gradient, local observable, depth 3 | Qubits 8 | PennyLane 38.7 ms | Qulacs 577 µs | Faster Qulacs 67.1× |
Gradient, local observable, depth 3 | Qubits 12 | PennyLane 82.6 ms | Qulacs 5.51 ms | Faster Qulacs 15.0× |
Gradient, local observable, depth 3 | Qubits 16 | PennyLane 462 ms | Qulacs 110 ms | Faster Qulacs 4.2× |
Statevector, global observable (all_z), depth 3 | Qubits 8 | PennyLane 6.89 ms | Qulacs 129 µs | Faster Qulacs 53.4× |
Statevector, global observable (all_z), depth 3 | Qubits 12 | PennyLane 12.1 ms | Qulacs 1.00 ms | Faster Qulacs 12.1× |
Statevector, global observable (all_z), depth 3 | Qubits 16 | PennyLane 62.8 ms | Qulacs 18.9 ms | Faster Qulacs 3.3× |
Strongly-entangling ansatz
RX/RY/RZ layers with a CNOT ring
| Workload | Qubits | PennyLane | Qulacs | Faster |
|---|---|---|---|---|
Expectation value, local observable, depth 3 | Qubits 4 | PennyLane 4.02 ms | Qulacs 68.2 µs | Faster Qulacs 59.0× |
Expectation value, local observable, depth 3 | Qubits 8 | PennyLane 7.98 ms | Qulacs 177 µs | Faster Qulacs 45.1× |
Expectation value, local observable, depth 3 | Qubits 12 | PennyLane 14.3 ms | Qulacs 1.56 ms | Faster Qulacs 9.1× |
Expectation value, local observable, depth 3 | Qubits 16 | PennyLane 70.1 ms | Qulacs 30.0 ms | Faster Qulacs 2.3× |
Expectation value, local observable, depth 3 | Qubits 20 | PennyLane 1.67 s | Qulacs 595 ms | Faster Qulacs 2.8× |
Expectation value, local observable, depth 3 | Qubits 24 | PennyLane skip | Qulacs 13.2 s | Faster – |
Expectation value, global observable (all_z), depth 3 | Qubits 4 | PennyLane 4.44 ms | Qulacs 77.7 µs | Faster Qulacs 57.1× |
Expectation value, global observable (all_z), depth 3 | Qubits 8 | PennyLane 8.95 ms | Qulacs 222 µs | Faster Qulacs 40.2× |
Expectation value, global observable (all_z), depth 3 | Qubits 12 | PennyLane 16.0 ms | Qulacs 2.50 ms | Faster Qulacs 6.4× |
Expectation value, global observable (all_z), depth 3 | Qubits 16 | PennyLane 78.2 ms | Qulacs 50.6 ms | Faster Qulacs 1.5× |
Expectation value, global observable (all_z), depth 3 | Qubits 20 | PennyLane 2.73 s | Qulacs 1.02 s | Faster Qulacs 2.7× |
Expectation value, global observable (all_z), depth 3 | Qubits 24 | PennyLane skip | Qulacs 21.4 s | Faster – |
Gradient, local observable, depth 3 | Qubits 4 | PennyLane 29.0 ms | Qulacs 278 µs | Faster Qulacs 104× |
Gradient, local observable, depth 3 | Qubits 8 | PennyLane 58.9 ms | Qulacs 822 µs | Faster Qulacs 71.6× |
Gradient, local observable, depth 3 | Qubits 12 | PennyLane 125 ms | Qulacs 8.63 ms | Faster Qulacs 14.5× |
Gradient, local observable, depth 3 | Qubits 16 | PennyLane 538 ms | Qulacs 180 ms | Faster Qulacs 3.0× |
Data re-uploading QML
Repeated encode-and-rotate classifier
| Workload | Qubits | PennyLane | Qulacs | Faster |
|---|---|---|---|---|
Expectation value, global observable (all_z), depth 2 | Qubits 8 | PennyLane 10.4 ms | Qulacs 185 µs | Faster Qulacs 56.2× |
Expectation value, global observable (all_z), depth 2 | Qubits 12 | PennyLane 18.2 ms | Qulacs 2.06 ms | Faster Qulacs 8.8× |
Expectation value, global observable (all_z), depth 2 | Qubits 16 | PennyLane 81.9 ms | Qulacs 41.8 ms | Faster Qulacs 2.0× |
Expectation value, global observable (all_z), depth 2, batch of 32 | Qubits 8 | PennyLane 14.7 ms | Qulacs 6.97 ms | Faster Qulacs 2.1× |
Expectation value, global observable (all_z), depth 2, batch of 32 | Qubits 12 | PennyLane 121 ms | Qulacs 67.1 ms | Faster Qulacs 1.8× |
Expectation value, global observable (all_z), depth 2, batch of 32 | Qubits 16 | PennyLane 8.25 s | Qulacs 1.33 s | Faster Qulacs 6.2× |
Expectation value, global observable (all_z), depth 2, batch of 256 | Qubits 8 | PennyLane 53.8 ms | Qulacs 48.4 ms | Faster Qulacs 1.1× |
Expectation value, global observable (all_z), depth 2, batch of 256 | Qubits 12 | PennyLane 1.08 s | Qulacs 529 ms | Faster Qulacs 2.0× |
Expectation value, global observable (all_z), depth 2, batch of 256 | Qubits 16 | PennyLane 71.1 s | Qulacs 10.6 s | Faster Qulacs 6.7× |
Expectation value, global observable (all_z), depth 2, batch of 1024 | Qubits 8 | PennyLane 200 ms | Qulacs 193 ms | Faster Qulacs 1.0× |
Expectation value, global observable (all_z), depth 2, batch of 1024 | Qubits 12 | PennyLane 13.7 s | Qulacs 2.13 s | Faster Qulacs 6.5× |
Expectation value, global observable (all_z), depth 2, batch of 1024 | Qubits 16 | PennyLane skip | Qulacs 42.7 s | Faster – |
Gradient, local observable, depth 2 | Qubits 8 | PennyLane unsupported | Qulacs 606 µs | Faster – |
Gradient, local observable, depth 2 | Qubits 12 | PennyLane unsupported | Qulacs 5.81 ms | Faster – |
Gradient, local observable, depth 2 | Qubits 16 | PennyLane unsupported | Qulacs 117 ms | Faster – |
Gradient, local observable, depth 2, batch of 32 | Qubits 8 | PennyLane unsupported | Qulacs 19.3 ms | Faster – |
Gradient, local observable, depth 2, batch of 32 | Qubits 12 | PennyLane unsupported | Qulacs 189 ms | Faster – |
Gradient, local observable, depth 2, batch of 32 | Qubits 16 | PennyLane unsupported | Qulacs 3.76 s | Faster – |
Gradient, local observable, depth 2, batch of 256 | Qubits 8 | PennyLane unsupported | Qulacs 153 ms | Faster – |
Gradient, local observable, depth 2, batch of 256 | Qubits 12 | PennyLane unsupported | Qulacs 1.47 s | Faster – |
Gradient, local observable, depth 2, batch of 256 | Qubits 16 | PennyLane skip | Qulacs 30.0 s | Faster – |
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.