A .zst file is a Zstandard-compressed stream. Zstandard hits roughly Gzip's ratio at several times the speed, and much better ratios when asked, which is why it has been quietly replacing Gzip across Linux distributions, databases and network protocols since about 2016.
Private for supported formats — processed in your browser
Convert supported files
Runs on your device
+
Drop ZST files here
Batch files can each use a different output. Nothing uploads for local conversions.
Working inputs include camera RAW, browser-local audio/video, PDF, CBZ/CBR comics, office documents, ebooks, markup, 3D models, structured text, images, and archives.
Yann Collet — who also wrote LZ4 — developed Zstandard at Facebook and released version 1.0 in 2016 under BSD and GPLv2. Adoption was fast for a compression format: the Linux kernel, btrfs, Arch Linux packages, Debian package compression, ZFS, RocksDB, and HTTP `Content-Encoding: zstd`, which browsers began shipping in 2024. RFC 8878 documents the format.
How ZST works
The frame magic is 28 B5 2F FD, followed by a frame header that may or may not declare the decompressed size — a decoder cannot assume it is there.
Compression combines a modern LZ77 matcher with finite state entropy (an implementation of asymmetric numeral systems), which is where the speed advantage over Gzip's Huffman coding comes from.
Levels run from 1 to 22, plus negative levels faster than LZ4; the format is identical at every level, so a low-level file needs no special decoder.
Trained dictionaries make small payloads compress dramatically better, and are the reason Zstandard suits many-small-records workloads.
Skippable frames let an application embed its own data in the stream without breaking a conforming decoder.
When to use ZST
Unpacking an Arch Linux or Debian package payload
Reading a compressed database backup or log archive
Opening a .tar.zst source release without installing zstd
Converting a compressed archive to ZIP for wider compatibility
Strengths and limitations
Strengths
Much faster than Gzip at a comparable ratio, and far faster to decompress
Scales from real-time speeds to ratios close to xz through a single format
An IETF-documented, permissively licensed standard with broad adoption
Limitations
Older tools still do not recognise it, so a .zst can be an obstacle rather than a convenience
The decompressed size is optional in the header, which complicates simple readers
Dictionary-compressed streams need the same dictionary to decode and are useless without it
Compatibility
The reference `zstd` tool, 7-Zip, PeaZip, GNU tar (`--zstd`) and bsdtar all handle it, and Chrome, Firefox and Edge accept zstd Content-Encoding. Novus Convert decodes .zst in the browser and rewrites the contents as ZIP, TAR or TAR.GZ. A stream holding a TAR is expanded to its members; one holding a single file is treated as that file, which is how a .zst is actually used in practice.
Zstandard carries no file metadata of its own; anything preserved comes from the TAR inside it.