Available nowRobust installerv0.4.0

QECTOR Decoder v3.

Source-available Rust/Python quantum error correction decoder library with Stim/Sinter workflows, PyMatching-compatible interfaces, belief-matching accuracy mode, BP-OSD / LDPC support, and CPU/GPU batch workflows.

Install with AIO installerGitHub Repository
Recommended install
git clone https://github.com/qectorlab/qector-decoder.git
cd qector-decoder

python install.py --install-rust
Installer

Use the AIO installer first.

The installer handles missing Rust, compatible Python selection, virtual environment setup, thin checkout repair, Git Bash linker path cleanup, prebuilt wheel fallback, dev dependencies, and pytest.

Open installer guideBenchmark evidence
Decoder stack

One library for validated QEC decoding workflows.

Rust core

Performance-sensitive decoder logic with Python bindings through PyO3.

Stim / Sinter

Detector Error Model loading and benchmark workflow integration.

PyMatching-compatible

Interfaces and comparison workflows for logical error rate validation.

Belief-matching

Accuracy mode for selected correlated workloads, slower than plain MWPM.

BP-OSD / LDPC

Support for LDPC and qLDPC experiments with reproducible reporting.

CPU/GPU batch

CPU, CUDA, and OpenCL batch workflows with tested CPU/GPU bit identity.

Validated scope

Honest public claims.

  • 832 tests collected: 829 passed / 2 skipped / 1 xfailed
  • Weighted MWPM matches PyMatching LER on tested circuit-level workloads
  • d=15 LER parity on tested memory workloads
  • Belief-matching shows lower observed LER on selected correlated workloads
  • BP-OSD is competitive on tested LDPC / qLDPC cases
  • PyMatching remains the latency leader for exact MWPM

Do not overclaim.

QECTOR should not be described as the universal fastest decoder or a universal PyMatching replacement. Its strengths are reproducibility, evidence packaging, accuracy-mode experimentation, BP-OSD / LDPC support, GPU batch workflows, and commercial integration.