packvium vs packingsolver, measured on OR-Library.
packingsolver against packvium on the pinned OR-Library profiles: Pareto-optimal in 3.
WHEN THIS WAS MEASURED · The sweep on this page ran against the 1.3.0 release on 2026-09-15, on a request set pinned on 2026-08-05. The raw artefacts per run are kept beside the baseline in benchmarks/results/COMPARATOR-BASELINE-1.3.0.md.
| Profile | packvium utilisation | packvium packed | packingsolver utilisation | packingsolver packed | Verdict for packingsolver |
|---|---|---|---|---|---|
| 3 item types · orlib-thpack-1 | 86.06% | 90.50% | 89.02% | 84.14% | Pareto-optimal |
| 12 item types · orlib-thpack-5 | 83.88% | 89.19% | 85.44% | 87.27% | Pareto-optimal |
| 20 item types · orlib-thpack-7 | 82.47% | 87.54% | 82.97% | 78.58% | Pareto-optimal |
Read from packingsolver's own public repository and documentation, not from running its code - unlike the benchmark numbers above, which this project's own validator recomputes on every run.
- Licence
- MIT
- Language
- C++
- Deployment
- A command-line tool and library built from source with CMake - no package-registry install for any language.
- Multiple containers per solve
- Yes
- Dependencies
- Pulls its solver dependencies from a vendored extern/ tree at build time, per its own CMake setup.
- Constraints modelled
- The broadest problem coverage of the six: 1D, 2D (rectangle and irregular polygon) and 3D box packing, including guillotine cuts, nesting and variable-sized bins.
- Determinism
- Its documentation does not state whether a given input always returns the same solution.
- Validation and refusals
- Reports waste and dimensional statistics per solve. Its documentation does not describe a reason for a specific item left unpacked.
- Limitations
- No published package for Python, PHP, Node or a browser runtime - using it means building the C++ source.
Two axes, and no blended score.
An engine is Pareto-optimal when nothing beat it on every axis at once. Runtime is recorded per run but is deliberately not an axis: one competitor only builds for linux/amd64 and runs under emulation here, so ranking that mixture would compare host arrangements rather than engines. Nothing on this page is a speed claim, and none should be read out of it.
See every engine on every profile