Apple Silicon MLX Accelerated · 1-25 Qubits · No CUDA Required

Quantum computing
on your Mac

High-performance quantum circuit simulation powered by MLX. Run QFT, VQE, and QCBM benchmarks on Apple Silicon with GPU acceleration. Democratizing quantum computing research.

M1 / M2 / M3 / M4 · State Vector Simulation · MLX GPU Acceleration · Open Source

Quantum Circuits

QFT, VQE, QCBM, Grover

MLX Engine

GPU + Unified Memory

Apple Silicon

M1 / M2 / M3 / M4 Max

QuantumStudio Interface

Run benchmarks interactively with our native macOS application. Real-time visualization, job management, and automatic result export.

PRX Quantum · In Preparation

mlxQuantum: High-Performance Quantum Circuit Simulation on Apple Silicon via Unified Memory Architecture

Abstract

We present mlxQ, an open-source quantum circuit simulation framework built on Apple's MLX array framework, targeting the unified memory architecture of Apple Silicon processors. By eliminating host-device memory transfers inherent in discrete GPU systems, mlxQ enables mid-scale state-vector simulations on consumer hardware with 128-512 GB unified memory. The framework provides a Python-first programming model with automatic Metal GPU acceleration, supporting single- and multi-qubit gate libraries, variational quantum algorithms (VQE, QAOA, QCBM), Hamiltonian simulation with Trotter-Suzuki decomposition, and OpenQASM 2.0 circuit import. We validate correctness through over 230 regression tests with analytical verification. On Apple M1 Max hardware, we demonstrate complete 25-qubit simulation for QFT (7.03s), QAOA (11.07s), and Hamiltonian evolution (40.73s), with sub-millisecond execution for shallow circuits.

1-25
Qubit Range
7.03s
QFT @ 25 Qubits
230+
Regression Tests
150+
Pages Documentation

Everything you need for quantum simulation

A complete toolkit for quantum computing research on Apple Silicon: state vector simulation, comprehensive gate library, algorithm implementations, and publication-ready benchmarks.

MLX Acceleration

Automatic GPU acceleration via Apple's MLX framework. Unified memory eliminates data transfer overhead between CPU and GPU.

GPU Native JIT Compile Metal
📈

Qubit Scaling

Simulate from 1 to 25 qubits with exponential scaling analysis. Sub-millisecond to hours of compute, all on consumer hardware.

1-25 Qubits State Vector MPS
🔧

Complete Gate Library

Full quantum gate support: Pauli, Hadamard, CNOT, Toffoli, phase gates, rotations, and custom unitaries with automatic adjoint.

60+ Gates Controlled Parametric
📊

Benchmark Suite

yaoquantum.org compatible benchmarks with PennyLane, Qulacs, and Yao.jl patterns. JSON/CSV export and automated plotting.

Publication Ready Plots CSV
25
Max Qubits
230+
Regression Tests
12
Tutorials
5
Frameworks Validated

Research-grade performance analysis

Run comprehensive quantum circuit benchmarks compatible with yaoquantum.org. Compare QFT, QCBM, VQE, and individual gate operations across qubit counts with automated visualization.

  • Quantum Fourier Transform (QFT): 7.03s at 25 qubits
  • QAOA ring optimization: 11.07s at 25 qubits
  • QCBM (9 layers): 26.28s at 25 qubits
  • Hamiltonian simulation: 40.73s at 25 qubits
# Run quantum benchmark suite ./bin64/mlxQuantumBenchmarks # Output: yaoquantum.org compatible { "benchmark": "qft", "qubits": [4, 8, 12, 16, 20, 25], "times_ms": [0.12, 0.89, 7.2, 142, 4820, 15360000], "backend": "mlx_gpu", "device": "M4 Max" } # Generates: qft_final.png, qcbm_final.png

Exponential Scaling Results

Complete benchmark scaling from 1-25 qubits on Apple M1 Max. All measurements confirm expected O(2n) complexity.

Unified memory advantage

Unlike discrete GPUs that require explicit memory transfers, Apple Silicon's unified memory architecture enables seamless CPU-GPU data sharing. This eliminates the primary bottleneck in quantum state manipulation.

  • M1 / M2 / M3 / M4 generation support
  • Up to 128GB unified memory for large states
  • 2.3x more energy efficient than CUDA alternatives
  • No CUDA installation or NVIDIA GPU required
M1 Max M1 Ultra M2 Pro M2 Max M3 Pro M3 Max M4 M4 Pro M4 Max

From gates to applications

Complete quantum algorithm implementations validated against analytical expectations and established frameworks. Run Grover's search, quantum phase estimation, variational circuits, and Hamiltonian simulation.

  • Grover's algorithm with oracle and diffusion
  • Quantum Fourier Transform (QFT)
  • Variational Quantum Eigensolver (VQE)
  • Phase estimation and Hamiltonian simulation
// Create quantum circuit auto circuit = Quantum::Circuit(4); circuit.H(0); circuit.CNOT(0, 1); circuit.RZ(1, theta); circuit.measure_all(); // Execute on MLX device auto device = Quantum::Device("mlx.gpu"); auto result = device.execute(circuit); // Results with timing telemetry std::cout << result.probabilities << std::endl; std::cout << "Time: " << result.exec_time_ms << "ms";

Publication-ready benchmarking

QuantumStudio is built on mlxQuantum, a research-grade quantum simulation framework targeting PRX Quantum publication. All benchmarks follow yaoquantum.org standards for direct comparison with established frameworks.

The framework has been validated against PennyLane, Qulacs, Yao.jl, and cuQuantum with comprehensive test coverage across 62 test cases spanning quantum operations, algorithms, and information-theoretic primitives.

PRX Quantum (In Preparation)
mlxQuantum: High-Performance Quantum Circuit Simulation on Apple Silicon via Unified Memory Architecture
Demonstrating research-grade quantum simulation on consumer hardware
We present mlxQuantum, a quantum circuit simulation framework leveraging Apple's MLX array framework for GPU-accelerated computation on Apple Silicon. Through comprehensive benchmarking across 1-25 qubits, we demonstrate that unified memory architecture enables practical quantum simulation with 2.3x energy efficiency compared to CUDA-based alternatives.

Built for serious quantum research

A production-ready platform for quantum computing research on Apple Silicon. Pure C++ core with MLX integration, Flutter desktop UI, and comprehensive benchmark compatibility.

MLX Framework

Apple's array framework for machine learning. Automatic GPU acceleration, just-in-time compilation, and unified memory optimization.

Flutter Desktop

Native desktop app for macOS. Material 3 design with responsive layouts, live benchmark logs, and result visualization.

Pure C++ Core

High-performance quantum simulation library with complete gate operations, state management, and algorithm implementations.

State Vector Simulation

Dense state vector representation with efficient tensor operations. Support for 1-25 qubits with exponential scaling analysis.

Benchmark Suite

yaoquantum.org compatible benchmarks with JSON/CSV export. Automated plotting with gnuplot integration.

Framework Compatibility

Validated against PennyLane, Qulacs, Yao.jl, and cuQuantum. API patterns match established quantum computing libraries.

Start simulating quantum circuits

Run research-grade quantum simulations on your Mac. Open source, Apple Silicon optimized, no CUDA required.

macOS · Apple Silicon · MLX · Open Source

The codebase is cross-platform, but we currently provide macOS binaries only. License: Source code is licensed under Business Source License 1.1 (BSL-1.1), and binary distributions are licensed under the osxQ Binary Distribution License. See LICENSE, BINARY-LICENSE.txt, and the website License page.