OpenZL

OpenZL is an open-source compression framework that generates specialized compressors optimized for specific data formats. Rather than using generic compression algorithms, it takes a description of your data structure and builds a custom compressor tailored to that format, achieving higher compression ratios while maintaining fast processing speeds. The project is designed for engineers handling large volumes of specialized datasets, particularly in AI workloads and datacenter pipelines.

OpenZL consists of a core library and code generation tools that produce format-specific compressors, all compatible with a single universal decompressor. According to the project, it delivers compression ratios significantly better than standard tools like zstd and xz while preserving the decompression speeds required by production systems. The framework is currently in active development and used extensively in production at Meta.