Benchmarks

Purrchiver matches or beats the standard command-line tools across essentially every format it writes — at its default settings, and there’s no format where it’s embarrassingly behind. Two machines below: an 8-P-core M2 Pro and a 4-P-core M4.

The wins come from a multithreaded engine:

• tar.xz — roughly 5–7× faster than the built-in tar. On a 170 MB mixed corpus: 8.2 s vs 56.8 s on the M2 Pro (~7×), 6.7 s vs 32.6 s on the M4 (~5×). Nothing else on macOS makes a fast tar.xz.
• Multi-file ZIP — parallel shard-and-merge. Faster than the built-in ZIP tools, single- or multi-file (M2 Pro: 1.91 s vs the built-in tools' 2.06–3.14 s; the M4 holds the same order at 1.48–2.44 s).
• 7z is at parity with the standard 7z tooling — same ratio, a hair faster (M2 Pro 6.33 s vs 6.89 s).
• Small-file folders pack leaner and faster than the built-in tar — no AppleDouble/xattr bloat (189 % overhead vs tar’s 441 %) — and at half the size of the system’s default zip.
• Streams hold parity with the standard CLIs; lzip is ~30–60% faster.

It’s a scaling story, not a single number. The biggest wins are on the multithreaded paths (tar.xz, 7z, parallel ZIP), where wall-clock tracks performance-core count and memory bandwidth — so the margins are larger on the 8-P-core M2 Pro than the 4-P-core M4, while the single-threaded stream formats land closer together. That’s why results are reported per machine: switch tabs for the full tables.

Honest caveats. Brotli isn’t apples-to-apples — Purrchiver compresses it at a fixed quality while the standard brotli CLI defaults to max quality 11, so that row isn’t a fair speed comparison. iso on large data favours the system’s built-in disk-image tool. And ZIP archives ≥ 4 GiB fall back to a correct single-threaded path (parallel Zip64 is deferred).