Official dataset release

The complete AIT Data Compression Challenge 2026 dataset consists of 16 files, A–P. Files A–H form the Training dataset and files I–P form the Testing dataset. The Testing dataset was hidden during the competition and is now publicly released for reproducibility.

Canonical files: use the individual files linked on this page for byte-for-byte reproduction of the benchmark. The data/data.zip is the Training archive containing A–H only and should not be interpreted as the complete A–P dataset.
8 files Training A–H · 13,136,308 bytes
8 files Testing I–P · 25,153,011 bytes
16 files Complete A–P · 38,289,319 bytes

Training dataset — A–H

Public during the competition. Total uncompressed size: 13,136,308 bytes.

FileBytesSHA-256Download
A 1,308,765 4847f6507db6987d1c31e0a692ccf40d1c8953d0c15feee8139615a6a1f7f8e7 Download A
B 1,168,767 e14f65d4ba595d0c7c84e21a43f07de7d1cac693220d7366517f2428f52051ab Download B
C 2,000,000 e53981006244ec5066a58e24939ffdafa7eab579b1fade14a9212662a8a5a354 Download C
D 2,000,000 dbb62ea0572cac060ea94f70b4562d7f1177143ce74fdce9eb25005df0044d43 Download D
E 1,012,112 ff07707d8705721e2ef9f0dc40fadf4af073cf17e072d26c7bab4188b1080e13 Download E
F 2,097,152 a4be1fcb6fca75616924fc9989dfea9d512a2972ca0465b4eca138574d47160f Download F
G 2,526,752 87ab967bcc7cba7192352915ecd147e9a96114923ea69bbc7fa32d7e7599ce16 Download G
H 1,022,760 7c5468b370f7c47eda07281e3437fafc568f95d10420051e3aa522709f9342c5 Download H

Testing dataset — I–P

Hidden until the end of the competition and used for the final evaluation. Total uncompressed size: 25,153,011 bytes.

FileBytesSHA-256Download
I 3,295,751 ce0c9f7822cb7cb7736c8e5916733f1334f4abc389c41c5f35d1daca42ac5bd2 Download I
J 3,000,000 7472a60ee841c2990407baed3b6820ce9bf94d85f00a4538227e075ec3da5d4e Download J
K 3,000,000 8073227627c5155d077fda27a7f372a577ece100dd252cf1c0cd814abbf80732 Download K
L 3,000,000 954ede81041d50edc458dadd9b7074519e1e97d957abf2f81ab2e81ff74d7509 Download L
M 3,145,728 f0b8d9f3f3b63f795bb40538703ba07454fa597a16c9936b1d0c5b483a2867df Download M
N 4,194,304 d787c7ca9ce500187efe980d7eeaa25f0fbca0240256b92871d2d7e999ccfa78 Download N
O 2,607,090 8a0ae94edb3bf4ee820b371e4f1d8ac8c47320e0442ef774c63a1f5874cd7524 Download O
P 2,910,138 da2ca4c742dc55216daeb68b6b7ce6d26fac0aae34f3b657c3322c14df4d0cc8 Download P

Integrity verification

Validate every downloaded file against the official SHA-256 catalog before reporting results.

Linux / macOS

sha256sum -c SHA256SUMS

Download the complete A–P dataset

mkdir -p AITDCC2026
cd AITDCC2026
for f in A B C D E F G H I J K L M N O P; do
  curl -fLO "https://aitdcc.github.io/data/$f"
done
curl -fLO "https://aitdcc.github.io/data/SHA256SUMS"
sha256sum -c SHA256SUMS

Download Testing only (I–P)

for f in I J K L M N O P; do
  curl -fLO "https://aitdcc.github.io/data/$f"
done

Reproducibility, commercial validation & attribution

The publicly released AITDCC 2026 files A–P may be used for independent reproducibility studies, lossless-compression benchmarking, scientific evaluation, and commercial-validation reports. Such use does not imply endorsement of any compressor, product, company, or reported result by the AITDCC organizers.

  • Mandatory attribution: any use of one or more AITDCC files must clearly identify the data as originating from the AIT Data Compression Challenge 2026 (AITDCC 2026) and link to the official benchmark page at https://aitdcc.github.io/.
  • Paper citation: publications, technical reports, benchmark reports, commercial-validation reports, presentations, and other public uses of the dataset must also cite the associated AITDCC paper: “The 2026 Algorithmic Information Theory Data Compression Challenge”.
  • Source provenance: when reporting results for individual files, users should retain and acknowledge the original provenance/source of the corresponding file whenever that provenance is provided by AITDCC or by the underlying source dataset.
  • Reproducibility: report the exact files evaluated (for example, A–H, I–P, or A–P) and verify them against the SHA-256 values published here.
  • Results: independently obtained results should be described as independent evaluations, not as official AITDCC competition results.
  • Third-party rights: where any underlying source material is subject to separate third-party attribution, citation, or licensing terms, those terms continue to apply.

In short, reuse is permitted for reproducibility, research, benchmarking, and commercial validation, provided that the AITDCC origin is credited, the associated AITDCC paper is cited, and any applicable third-party source terms are respected.