Deepnest, deepnest-next, SVGnest or NestForge?

Full disclosure first: we build NestForge, so we are not neutral. What we can offer instead of neutrality is checkable facts: release dates from the repositories, feature claims you can verify in five minutes, and benchmark numbers with seeds, including the instances where NestForge loses.

Status and basics, verified

Dates and issue counts come from the public GitHub repositories, checked July 2026. If you read this later, re-check them; the deepnest-next project moves.

Where the free tools are the better choice

Honesty cuts both ways, so here is where we would not recommend NestForge. If you cut plasma and want fewer pierces through shared edges, Deepnest's common-line merging is the feature that matters, and NestForge does not have it. If your source files are CorelDRAW CDR, the Deepnest family imports them directly. If you nest bitmap images for engraving, same story. And if your budget is zero with more than 10 parts per sheet, deepnest-next is free without limits. It is a genuinely active community project and we say that without an asterisk.

The benchmark, losses included

We ran NestForge, SVGnest (the live site) and deepnest-next v1.5.6 on 12 classic ESICUP nesting instances. Fairness rules: the competitors got at least as much time as our slowest run, minimum 60 seconds, and deepnest-next used its 4 threads while NestForge ran single-threaded. Strip length decides; smaller is better.

Result: NestForge produced the shortest strip on 9 of 12 instances against both tools. It beat deepnest-next on all 12. Against SVGnest it lost twice, and we would rather you hear that from us: on jakobs2, SVGnest reached 27.4 against our 28.0, and on trousers 241.9 against our 249.4. On jakobs1 both hit exactly 13.0. The full per-instance table with seeds, times and utilization is on the benchmarks section of the main page.

One more layer, because it is the core difference in philosophy: we re-checked every layout from all three tools with the same independent exact-arithmetic validator that gates NestForge's own results. All 72 NestForge runs pass. Of the competitor layouts, 2 of 24 SVGnest runs pass and 0 of 24 deepnest-next runs do. The violations are tiny, micrometre scale, and in practice they usually do not matter for cutting. We counted the competitors' lengths at full value anyway. But if you have ever had a nester return parts that overlap for real (a known, long-standing Deepnest issue), you know why we made the validator non-negotiable.

Broken files: named errors vs. silence

We also feed all three tools deliberately broken files: NaN coordinates, truncated exports, self-intersecting contours, open paths, empty files. NestForge answers each one with a named error that says what is wrong and where. SVGnest reacts with unhandled JavaScript exceptions or imports the broken geometry silently. deepnest-next mostly imports "successfully" with zero parts and no message, and for DXF it depends on a cloud conversion server that can fail with a plain HTTP 500. The exact file-by-file protocol is public in our repository documentation, and you can reproduce it with the files from the corpus.

How to choose

Pick deepnest-next if free and common-line matter more than reproducibility. Pick SVGnest for occasional small SVG jobs with zero setup. Pick NestForge if your files are real-world DXF exports with real problems, if you need the same layout twice, or if a layout that is mathematically guaranteed not to overlap is worth a one-time purchase to you. Try the free tier on your own files first; that is what it is for.