GoKawiilQumulator SDK
Simulate 1,000-qubit quantum circuits on classical hardware. Exact results. No GPU. No quantum hardware required.
What is this?
Qumulator is a cloud API — and this is its Python client — for simulating quantum circuits, spin systems, photonic amplitudes, and molecular properties on standard classical hardware. It does not require a quantum computer, a GPU, or any special hardware. It runs in the cloud (Google Cloud Run, 4 vCPU, 16 GB RAM) and returns results over HTTP.
The key numbers: a 1,000-qubit circuit at depth 3 runs in under 1 second using 1 MB of memory — where the equivalent statevector would require $2^{1000}$ bytes (more atoms than exist in the observable universe). A 105-qubit Willow-layout circuit at depth 5 completes in under 0.5 s. Results are exact within the stated truncation error, not statistical estimates.
The simulation engine is built on the KLT Engine, a proprietary classical simulation framework that routes each problem to the most efficient representation — tensor network, cluster-exact, Gaussian covariance matrix, nexus graph, or full statevector — based on the entanglement structure of the specific circuit. Callers select a mode with a single parameter; the engine handles routing automatically.
Benchmarks
Measured on a standard cloud CPU (4 vCPU, no GPU). "Exact" means output agrees with full statevector simulation to double-precision floating point (< 10⁻¹⁴ L² error on the amplitude vector).
Problem Size Result Reference Error Time CHSH Bell violation N=2 S = 2.828427 2√2 = 2.828427 < 0.0001% < 1 ms H₁₂ Heisenberg chain 12 sites −11.000 −11.000 (exact diag.) 0.00% ~0.27 s Photonic hafnian (GBS) 8×8 matrix 0.2598−0.0078i exact DP < 2×10⁻¹⁵ 39 ms Photonic hafnian (GBS) 12×12 matrix 0.0239+0.9947i exact DP < 5×10⁻¹⁵ 43 ms RCS circuit (exact) 12 q, depth 20 XEB = 1.014 exact statevector 0.00% 15–23 ms RCS circuit (exact) 20 q, depth 20 XEB = 1.024 exact statevector 0.00% 8.5–9.6 s MBL discrete time crystal 8 q, 24 Floquet autocorr = 0.827 Google Sycamore 2021 Consistent ~1 s Holographic wormhole 2×6 SYK sites fidelity 94.89% Google Sycamore 2022 — ~5 s Non-Abelian anyon braiding Fibonacci anyons ‖[σ₁,σ₂]‖ = 1.272 SU(2)₃ exact < 0.001% < 1 ms Kitaev chain BdG L=1000 sites W=−1, gap=2.000 analytic (exact) < 10⁻¹² 0.84 s QUBO dense optimisation N=100 matches SA optimum simulated annealing 0% ~3 s Kuramoto BEC (large-scale) N=500 oscillators, 2 MB r=0.114 (Mott-like) statevector: 2⁵⁰⁰ bytes (impossible) — 3.22 s
Circuit depth limits (approximation modes)
... continue reading