facebook/openzl

Health v2 · maintenance only

Flags: no_license

Adoption not part of the score

3272 stars · 174 forks observed · 2026-10-15

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

OpenZL is a format-aware lossless data compression framework from Meta that builds specialized compressors from a description of your data format, composed as DAGs of codecs. It ships a core C library, a zli CLI, training tools, and a universal decompressor compatible with all generated compressors.

Use cases

• compress large numeric arrays faster and smaller than zstd
• compress structured data like csv or json with format-aware parsing
• train a specialized compressor for my custom binary data format
• compress AI/ML training data at datacenter speeds
• build a custom compression pipeline from composable codecs
• benchmark compression tools on my own files

When to choose

• you handle large volumes of specialized, structured, or numeric datasets and need both high ratios and high throughput
• you want one universal decompressor to read frames produced by many different specialized compressors
• you need production-grade compression used at Meta scale with forward-compatibility guarantees

When to avoid

• you just need simple general-purpose compression of unstructured files with a mature ecosystem
• you need a frozen, stable API and compressed format - both are still evolving
• you rely on custom codecs, which break compatibility with the standard OpenZL decompressor

Facets

library · maturity active

compression serialization cli developer-tools files developer-tools performance windows cross-platform cli python cpp c data-compression format-aware-compression lossless-compression compression-graphs universal-decompressor zstd sddl training-tool data-engineering linux macos

Member repositories

For agents

markdown · JSON · MCP: product_card(name="facebook/openzl")