OVERWRLD's program altxs 1.0.0 reached #1 on Matt Mahoney's Large Text Compression Benchmark (LTCB), the field's hardest widely cited public board for lossless text compression.
On the board as listed 6 September 2026, altxs 1.0.0 compressed enwik9, the first 1,000,000,000 bytes of a fixed English Wikipedia dump, to an official total size of 106,924,811 bytes. That is 93,434,410 bytes of payload plus 13,490,401 bytes of decompressor. LTCB ranks by that total: the compressed file plus a zip of whatever is needed to read it back. The decoder cannot be left off the score.
The next listed total on the same ranking is nncp v3.2 at 107,261,318 bytes.
Why the rules matter
Many compression claims leave the model or decoder out of the size. LTCB does not; its rules count the decompressor on purpose. The public row is the authority for our rank claim. A later 91.7 MB model is not on the board and is not the rank.
Two proofs, kept separate
- The benchmark. altxs 1.0.0 moved the public floor on enwik9 under LTCB rules.
- The product. A portable lossless compressor built to keep most of that ratio, with decode cheap enough to ship where data already lives.
Our commercial focus is video first, then AI infrastructure and data-center workloads, where storage, bandwidth, compute, and energy are each paid for the same bytes. See Compression is capacity.
The numbers
| Field | Value |
|---|---|
| Program | altxs 1.0.0 |
| Corpus | enwik9 (1,000,000,000 bytes in) |
| Mode | lossless, decompressor counted |
| Payload | 93,434,410 bytes |
| Decompressor | 13,490,401 bytes |
| Official total (rank metric) | 106,924,811 bytes |
| Public rank | #1 |
| Submitted | 2026-08-25 |
| Listed | 2026-09-06 |
| Next listed total | nncp v3.2, 107,261,318 bytes |
Media and partner inquiries: hello@overwrld.ai