Compression codec benchmark

Each matrix compares six codecs across the same five metrics. Cell labels show measured values; background intensity ranks codecs within that dataset.

Reading the matrix: darker cells are better. Memory is inverted because lower usage is preferable. Results remain specific to this run's hardware, software, datasets, and settings.

RatioCompression ratio · higher is better
Space savedSpace saved · higher is better
CompressCompression throughput · higher is better
DecompressDecompression throughput · higher is better
MemoryPeak RSS · lower is better

Run environment and methodology

CPUAMD64 Family 26 Model 96 Stepping 0, AuthenticAMD

Cores8 physical / 16 logical

Memory31.14 GiB

PlatformWindows-11-10.0.26200-SP0 · AMD64

Python3.12.10

Run2026-07-28 09:15:20.512845 UTC · profile standard

Method5 measured + 1 warm-up · 1.0 MiB chunks · 1 process

Input2.43 GiB across 4 datasets

enwik9 · Wikipedia text codec comparison matrix

953.7 MiB · Median across 5 measured repetitions

GH Archive · JSON events codec comparison matrix

977.8 MiB · Median across 5 measured repetitions

LibriSpeech test-clean · FLAC audio codec comparison matrix

350.2 MiB · Median across 5 measured repetitions

Silesia · mixed file corpus codec comparison matrix

202.1 MiB · Median across 12 files