惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

D
DataBreaches.Net
IT之家
IT之家
The Cloudflare Blog
Apple Machine Learning Research
Apple Machine Learning Research
WordPress大学
WordPress大学
N
Netflix TechBlog - Medium
阮一峰的网络日志
阮一峰的网络日志
P
Proofpoint News Feed
L
LangChain Blog
博客园 - Franky
美团技术团队
J
Java Code Geeks
Microsoft Security Blog
Microsoft Security Blog
博客园 - 叶小钗
小众软件
小众软件
Y
Y Combinator Blog
B
Blog RSS Feed
Cyber Security Advisories - MS-ISAC
Cyber Security Advisories - MS-ISAC
D
Docker
Hugging Face - Blog
Hugging Face - Blog
Jina AI
Jina AI
罗磊的独立博客
大猫的无限游戏
大猫的无限游戏
Vercel News
Vercel News

Hacker News

GitHub - SeanFDZ/macmind: Single-layer transformer in HyperTalk for the classic Macintosh Show HN: Agent-cache – Multi-tier LLM/tool/session caching for Valkey and Redis Bonsai 1-bit WebGPU - a Hugging Face Space by webml-community Moving a large-scale metrics pipeline from StatsD to OpenTelemetry / Prometheus GitHub - Nightmare-Eclipse/RedSun: The Red Sun vulnerability repository GitHub - SethPyle376/hiraeth: Local AWS emulator focused on fast integration testing, with SQS support, SQLite-backed state, and a debug-friendly web UI. GitHub - macOS26/Agent: Any AI, replaces Claude Code, Cursor, OpenClaw. Over 18 LLM providers (Claude, OpenAI, Gemini, Ollama, Zai, HF, Qwen) wired into a native Mac app that writes code, builds Xcode projects, bumps versions, manages git, automates Safari, use AppleScript, JS or Accessibility, extend Agent! w/ MCP Servers, run tasks from your iPhone via Messages. YouTube now lets you turn off Shorts I Made a Terminal Pager Burgers | マクドナルド公式 Commands — HackerNews CLI documentation ChatGPT for Excel PiCore - Raspberry Pi Port of Tiny Core Linux Live Nation illegally monopolized ticketing market, jury finds Google Broke Its Promise to Me. Now ICE Has My Data. Founding Engineer at Adaptional | Y Combinator CRISPR takes important step toward silencing Down syndrome’s extra chromosome GitHub - saffron-health/libretto: The AI toolkit for building reliable browser automations US v. Heppner (S.D.N.Y. 2026) no attorney-client privilege for AI chats [pdf] Retrofitting JIT Compilers into C Interpreters IPv6 – Google The Accursèd Alphabetical Clock Cybersecurity Looks Like Proof of Work Now Fragments: April 14 Cal.com Goes Closed Source: Why AI Security Is Forcing Our Decision | Cal.com - Scheduling Software for Online Bookings Laravel raised money and now injects ads directly into your agent When moving fast, talking is the first thing to break Too much Discussion of the XOR swap trick – Heather Cafe Introduction to Spherical Harmonics for Graphics Programmers The Grand Line
GitHub - qumulator/qumulator-sdk: Python client SDK for t...
nnoorbakhsh · 2026-04-27 · via Hacker News

Test SDK PyPI version Python 3.10+ MIT License

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)

Bond dimension $\chi = 2^\text{depth}$; all tiers keep peak memory under 400 MB.

Tier Qubit range Max depth Peak memory Notes
1 1 – 20 20 335 MB Exact for structured circuits
2 21 – 54 6 3.5 MB χ = 64 (exact)
3 55 – 105 5 1.7 MB χ = 32 (exact)
4 106 – 1,000 3 1 MB χ = 8 (exact)

Statevector mode: max 20 qubits at any depth.


Install

pip install qumulator-sdk

Optional extras:

pip install "qumulator-sdk[qiskit]"   # Qiskit drop-in backend
pip install "qumulator-sdk[cirq]"     # Cirq drop-in simulator
pip install "qumulator-sdk[all]"      # everything

Get a free API key

curl -s -X POST https://api.qumulator.com/keys \
     -H "Content-Type: application/json" \
     -d '{"name": "my-key"}' | python -m json.tool

Or via the CLI (after install):

qumulator key

Quick start — 1,000-qubit circuit in 30 seconds

import os
from qumulator import QumulatorClient

client = QumulatorClient(
    api_url="https://api.qumulator.com",
    api_key=os.environ["QUMULATOR_API_KEY"],
)

# 1000-qubit GHZ state via the fluent builder API
eng = client.circuit.engine(n_qubits=1000)
eng.apply("h", 0)
for i in range(999):
    eng.apply("cx", [i, i + 1])

result = eng.sample(shots=1024)
print(result)
# {'0000...0000': 512, '1111...1111': 512}
# Exact result. No quantum hardware. No GPU. Standard cloud CPU.

Run the built-in demo against the live API:

qumulator demo           # 1000-qubit GHZ
qumulator demo --willow  # 105-qubit Willow-layout RCS

OpenQASM 2/3

result = client.circuit.run_qasm("""
OPENQASM 2.0;
include "qelib1.inc";
qreg q[2];
creg c[2];
h q[0];
cx q[0], q[1];
measure q -> c;
""", shots=1024)

print(result.counts)       # {'00': ~512, '11': ~512}
print(result.entropy_map)  # [0.999, 0.999] — entanglement per qubit

Drop into Qiskit — two lines of code

from qumulator.backends.qiskit_backend import QumulatorBackend
from qiskit import transpile

backend = QumulatorBackend(client)               # replaces AerSimulator()
job     = backend.run(transpile(qc, backend), shots=1024)
counts  = job.result().get_counts()

Everything else in your Qiskit workflow is unchanged.


Drop into Cirq — two lines of code

from qumulator.backends.cirq_simulator import QumulatorSimulator

sim    = QumulatorSimulator(client)              # replaces cirq.Simulator()
result = sim.run(circuit, repetitions=1024)

Simulation modes

Pass mode= to any run() call. The server selects auto by default.

User-facing mode Internal mode Max qubits Best for
auto (server-routed) 1,000 General circuits; server auto-routes
exact (statevector) 20 Unconditionally correct; small N
compressed (tensor network) 1,000 VQE, QAOA, chemistry ansätze
tensor (MPS) 1,000 1D-structured, low-entanglement circuits
hamiltonian (operator algebra) 1,000 Hamiltonian simulation without gate decomposition
gaussian (covariance matrix) unlimited Clifford circuits; returns Wigner negativity certificate

Other computation types

# Molecular HOMO/LUMO frontier orbital energies (SMILES input)
homo = client.homo.run("Oc1ccc(/C=C/c2cc(O)cc(O)c2)cc1")
print(homo.homo_E_eV, homo.lumo_E_eV, homo.gap_eV)

# Ground-state energy of a spin Hamiltonian (Ising / Heisenberg / general)
import numpy as np
J = np.random.randn(8, 8); J = (J + J.T) / 2
result = client.klt.run(J.tolist())
print(result.energy)

# Hafnian / GBS photonic amplitude
A = np.random.randn(8, 8); A = (A + A.T) / 2
h = client.hafnian.run(A.tolist())
print(h.value)

Free tier limits

Limit Value
Compute Units / month 500 CU (1 CU = 1 CPU-second of engine time)
Max qubits (statevector mode) 20
Max qubits (MPS mode) 54
Max depth (MPS, ≤ 54 q) 9
Rate limit 1 request / minute
Daily limit 100 requests / day
Free tier availability Beta only — may be discontinued at any time

Paid plans start at $99/month (10,000 CU). See qumulator.com/#pricing.


Demo notebooks

Click to open in Google Colab — no install required, just add your API key:

Notebook Description
Open in Colab Willow RCS — 105-qubit exact simulation, Willow-layout
Open in Colab Holographic wormhole — traversable wormhole, matches Google 2022
Open in Colab Anyon braiding — Fibonacci anyons, matches Microsoft topological target
Open in Colab Discrete time crystal — MBL Floquet, matches Google Sycamore 2021
Open in Colab QUBO optimisation — 100-variable dense combinatorial optimisation

CLI

The qumulator command ships with the SDK:

qumulator demo               # 1000-qubit GHZ demo vs. the public API
qumulator demo --willow      # 105-qubit Willow-layout benchmark
qumulator demo --wormhole    # holographic wormhole
qumulator demo --anyon       # anyon braiding
qumulator key                # instructions to get a free API key
qumulator run circuit.qasm   # submit a QASM file and print the result

Set QUMULATOR_API_KEY in your environment, or pass --key YOUR_KEY.


Documentation

Full API reference and examples: qumulator.com


License

MIT — see LICENSE.