Today we're announcing PZIP, a fundamentally new approach to data compression. While traditional compressors like LZMA, gzip, and zstd treat all data as streams of bytes, PZIP understands what your data actually is.

The Problem

A CSV file has columns and types. A log file has repeating templates. Your JSON has a discoverable schema. But your compressor ignores all of it. It sees bytes, not structure.

That's like trying to summarize a book by counting letter frequencies instead of understanding the plot.

The Results

We tested PZIP against LZMA-9 (maximum compression) on 3,184 real-world files across 20 file types:

3,117 wins — PZIP beats the best competitor on 98% of all files
Up to 93.5% smaller than LZMA on structured data
Byte-exact round-trip verified on every single file

How It Works

PZIP uses 151 specialized compression strategies — we call them weapons. Each one targets a specific data pattern: timestamps, floating point numbers, dictionaries, templates, sequences, and more.

The key insight: don't compress the data. Compress the generator. If your CSV column contains sequential IDs from 1 to 10,000, PZIP stores three numbers (start=1, step=1, count=10000) instead of 10,000 integers.

Never-Worse Guarantee

If PZIP can't beat LZMA on a file, it simply outputs LZMA. You literally cannot lose. This is our never-worse guarantee, verified on every operation.

Try It Now

PZIP is free during beta. Upload any file up to 30 MB at pzip.net/demo and see the difference for yourself. No signup, no credit card, no data stored.

For enterprise deployments (Python SDK, REST API, on-prem), contact us.