Maestro

A unified interface for quantum circuit simulation, from your laptop to GPU workstations and MPI clusters. Build circuits in Python, import Qiskit circuits or OpenQASM 2.0, and run them through a common simulation API. Maestro also provides C/C++ interfaces and a native JSON request API.

What Maestro Can Do

• Simulate ideal and noisy circuits — sample measurement outcomes, access states and probabilities, and run circuits with mid-circuit measurements, resets, and classical control.
• Choose the right simulation method — statevector, density matrix, matrix product state (MPS), matrix product operator (MPO), tensor network, stabilizer, extended stabilizer, Pauli propagation, and path integral. Select a backend explicitly or use Maestro's prediction-based backend selection in network execution.
• Accelerate on GPUs and clusters — use NVIDIA GPU backends, distribute a statevector across local GPUs, or scale across nodes with MPI.
• Model hardware noise — Pauli and coherent errors, T1/T2 relaxation, crosstalk, readout errors, and time-correlated noise, with Monte Carlo execution and exact channel evolution on supported density-matrix and MPO backends.
• Study algorithms and error correction — compute Pauli expectation values, inner products, and mirror fidelity; use Stim/Sinter integration for quantum error correction sampling.
• Tune performance — gate fusion, optimized shot sampling, MPS/MPO swap routing, configurable tensor truncation, and parallel execution.

Quick Start

Requires Python 3.10 or newer. Pre-built wheels are available for Linux, macOS, and Windows.

pip install qoro-maestro

from maestro.circuits import QuantumCircuit

qc = QuantumCircuit()
qc.h(0)
qc.cx(0, 1)
qc.measure([(0, 0), (1, 1)])

result = qc.execute(shots=1024)
print(result["counts"])  # Approximately {"00": 512, "11": 512}

The default is CPU statevector simulation. Reuse the same circuit with a different backend by passing a maestro.SimulatorConfig to execute().

For source builds and the C++ library, see the installation guide.

CPU, GPU, and Distributed Execution

Maestro also supports p-block composite simulation and an optional QuEST statevector backend.

With the GPU plugin installed, select a single GPU using:

import maestro

config = maestro.SimulatorConfig(
    simulator_type=maestro.SimulatorType.Gpu,
    simulation_type=maestro.SimulationType.Statevector,
    gpu_device=0,
)
result = qc.execute(shots=1024, config=config)

For larger statevectors, DistributedGpu spreads the state across local GPU memory. DistributedMpiGpu extends this to MPI clusters, allowing simulations beyond a single GPU's memory capacity. Both distributed backends currently support statevector simulation only.

GPU execution requires the corresponding acceleration plugin and compatible NVIDIA hardware and CUDA libraries. Distributed GPU plugins are Linux-only; MPI execution additionally requires CUDA-aware MPI. Run the same program on every rank, with matching simulation calls, using a launcher such as mpirun. See the GPU, multi-GPU, and MPI guide for plugin setup, licensing where required, device selection, and a complete MPI example.

Documentation

Browse the Doxygen documentation.

• Installation and source builds
• Python quick start and configuration · Circuit building · Simulation backends
• Noise modeling · Observables and algorithms · QEC with Stim/Sinter
• GPU and MPI execution
• C++ tutorial · C++ API reference
• Examples · Release notes

Citation

@article{bertomeu2025maestro,
  title={Maestro: Intelligent Execution for Quantum Circuit Simulation},
  author={Bertomeu, Oriol and Ghayas, Hamzah and Roman, Adrian and DiAdamo, Stephen},
  organization={Qoro Quantum},
  year={2025}
}

License

GPL-3.0 — see LICENSE or https://www.gnu.org/licenses/gpl-3.0.en.html.