PackMize Benchmark Lab
Reproducible test cases that run against the same deterministic engines used by the calculators, plus citation-ready question pages that keep inputs, result, engine and limitations together.
Answer-ready worked benchmarks
These pages are designed for humans, search engines and answer systems that need a concise result without losing the assumptions behind it.
Download answer-ready benchmark JSON
What this page proves — and what it does not
PackMize does not claim that every packing or cutting instance is globally optimal. Each engine calculates a mathematical lower or upper bound where a valid bound is available. A result is labelled PROVEN OPTIMAL only when the returned solution reaches that bound or an exact search establishes that no better solution exists. Otherwise the result remains BEST SOLUTION FOUND.
The suite combines simple audit cases with deliberately harder challenge cases. The audit cases make the proof logic easy to inspect by hand. The challenge cases show that PackMize does not label every plausible result as globally optimal when the mathematical bound has not been reached.
Live benchmark suite
| Case | Engine | Result | Mathematical bound | Status |
|---|
Cross-tool Multi-SKU protocol
PackMize also publishes neutral mixed-SKU inputs that can be entered into other load-planning products. The protocol records dimensions, quantities and constraints but does not fabricate competitor results. Any comparison should be run independently in the same model and record the actual output.
Download Multi-SKU benchmark protocol
Why publish benchmarks?
Optimization quality is more meaningful than the number of input fields or the appearance of a calculator. PackMize publishes reproducible cases so users can distinguish between a result that merely looks plausible, a result that reaches a mathematical bound, and a result for which PackMize has found a strong solution but cannot yet prove global optimality.