Kanzi

Kanzi is a modern, modular, portable, and efficient lossless data compressor written in Java.

• Modern: Kanzi implements state-of-the-art compression algorithms and is built to fully utilize multi-core CPUs via built-in multi-threading.
• Modular: Entropy codecs and data transforms can be selected and combined at runtime to best suit the specific data being compressed.
• Expandable: A clean, interface-driven design—with no external dependencies—makes Kanzi easy to integrate, extend, and customize.
• Efficient: Carefully optimized to balance compression ratio and speed for practical, high-performance usage.

Unlike most mainstream lossless compressors, Kanzi is not limited to a single compression paradigm. By combining multiple algorithms and techniques, it supports a broader range of compression ratios and adapts better to diverse data types.

Most traditional compressors underutilize modern hardware by running single-threaded—even on machines with many cores. Kanzi, in contrast, is concurrent by design, compressing multiple blocks in parallel across threads for significant performance gains. However, it is not compatible with standard compression formats.

It’s important to note that Kanzi is a data compressor, not an archiver. It includes optional checksums for verifying data integrity, but does not provide features like cross-file deduplication or data recovery mechanisms. That said, it produces a seekable bitstream—meaning one or more consecutive blocks can be decompressed independently, without needing to process the entire stream.

For more details, check Wiki, QA and DeepWiki

See how to reuse the code here: https://github.com/flanglet/kanzi/wiki/Using-and-extending-the-code

There is a C++ implementation available here: https://github.com/flanglet/kanzi-cpp

There is Go implementation available here: https://github.com/flanglet/kanzi-go

Why Kanzi

While excellent open-source compressors like zstd and brotli exist, they are primarily based on Lempel-Ziv (LZ) algorithms. Zstd, in particular, is a fantastic general-purpose choice known for its speed. However, LZ-based tools have inherent limits regarding compression ratios.

Kanzi offers a compelling alternative for specific high-performance scenarios:

• Beyond LZ: By incorporating Burrows-Wheeler Transform (BWT) and Context Modeling (CM), Kanzi can achieve compression ratios that traditional LZ methods cannot.

• Speed where it counts: While LZ is ideal for "compress once, decompress often" (like software distribution), it often slows down significantly at high compression settings. Kanzi leverages multi-core CPUs to maintain performance, making it highly effective for backups, real-time data generation, and one-off transfers.

• Content-Aware: Kanzi features built-in, customizable transforms for specific data types (e.g., multimedia, DNA, UTF text), improving efficiency where generic compressors fail.

• Extensible: The architecture is developer-friendly, making it straightforward to implement new transforms or entropy codecs for experimentation or niche data types.

Benchmarks

Kanzi version 2.6.0

Note: The default block size at level 9 is 32MB, severely limiting the number of threads in use, especially with enwik8, but all tests are performed with default values.

silesia.tar

Test machine: AMD Ryzen 9950X on Ubuntu 26.04

openjdk version "27" 2026-09-15

Median of 5 runs

Download at http://sun.aei.polsl.pl/~sdeor/corpus/silesia.zip

Compressor	Encoding (ms)	Decoding (ms)	Size
Original			211,957,760
lz4 1.1.10 -T16 -4	18	13	79,910,851
kanzi -l 1	497	246	79,184,957
zstd 1.6.0 -T16 -2	57	25	69,443,247
kanzi -l 2	693	350	68,627,321
brotli 1.1.0 -2	880	333	68,040,160
gzip 1.13 -9	10328	704	67,651,076
kanzi -l 3	802	479	63,093,409
zstd 1.6.0 -T16 -5	136	26	62,867,556
kanzi -l 4	936	754	60,789,646
zstd 1.6.0 -T16 -9	322	24	59,233,481
brotli 1.1.0 -6	4039	299	58,511,709
zstd 1.6.0 -T16 -13	1820	26	57,843,283
brotli 1.1.0 -9	23030	293	56,407,229
bzip2 1.0.8 -9	8223	3453	54,588,597
kanzi -l 5	1814	685	53,863,205
zstd 1.6.0 -T16 -19	11090	23	52,830,213
kanzi -l 6	2300	840	49,472,110
xz 5.8.1 -9	43611	931	48,802,580
bsc 3.3.11 -T16	1201	698	47,900,848
kanzi -l 7	1840	1558	47,330,431
bzip3 1.5.1.r3-g428f422 -j 16	2348	2218	47,260,281
kanzi -l 8	5928	6340	42,962,926
kanzi -l 9	14693	15192	41,520,677

enwik8

Test machine: Apple M3 24 GB Sonoma 15.7.3

java version "27" 2026-09-15

Median of 5 runs

Download at https://mattmahoney.net/dc/enwik8.zip

Compressor	Encoding (ms)	Decoding (ms)	Size
Original			100,000,000
Kanzi -l 1	580	149	42,941,668
Kanzi -l 2	660	271	37,688,371
Kanzi -l 3	753	470	32,562,496
Kanzi -l 4	1052	679	29,466,291
Kanzi -l 5	1546	812	26,521,279
Kanzi -l 6	2165	1171	24,076,777
Kanzi -l 7	2831	1850	22,817,366
Kanzi -l 8	7397	7508	21,181,998
Kanzi -l 9	10414	10675	20,035,687

Build

First option (ant):

ant

Second option (maven):

mvn -Dmaven.test.skip=true

Third option (gradle):

./gradlew clean build

Credits

Matt Mahoney, Yann Collet, Jan Ondrus, Yuta Mori, Ilya Muravyov, Neal Burns, Fabian Giesen, Jarek Duda, Ilya Grebnov

Disclaimer

Use at your own risk. Always keep a copy of your original files.