Dolma

license

ODC-BY

ungated

yes

dataset_card

yes

code

apache-2.0-toolkit

Permissively licensed (ODC-BY), ungated download with a full dataset card and replication toolkit.

• https://huggingface.co/api/datasets/allenai/dolma recorded 2026-09-29

  Hub API reports the dataset as not gated and ODC-BY licensed, with tags "license:odc-by" and the dataset description "Dolma: an Open Corpus of Three Trillion Tokens for Language Model Pretraining Research".

• https://huggingface.co/datasets/allenai/dolma recorded 2026-09-29

  Hugging Face dataset page for allenai/dolma renders the card ("Dolma is a dataset of 3 trillion tokens from a diverse mix of web content, academic publications, code, books, and encyclopedic materials") with the license tag "odc-by".

• https://huggingface.co/datasets/allenai/dolma/raw/main/README.md recorded 2026-09-29

  Dataset card front matter declares "license: odc-by" and the card describes "Dolma is a dataset of 3 trillion tokens from a diverse mix of web content, academic publications, code, books, and encyclopedic materials."

Adoption is measured by Hugging Face downloads of the dataset over the trailing month, the only usage channel on record for Dolma.

• https://huggingface.co/api/datasets/allenai/dolma recorded 2026-08-13

  2,724 downloads in the trailing 30 days for allenai/dolma

Dolma trained OLMo 1/2 and OLMoE, but the FineWeb paper's own ablation ranks it behind FineWeb, and Ai2 has since superseded it with Dolma 3.

• https://allenai.org/blog/olmo3 recorded 2026-09-29

  Ai2 Olmo 3 announcement says "Olmo 3 is pretrained on" Dolma 3, "a new ~9.3-trillion-token corpus".

• https://arxiv.org/html/2406.17557 recorded 2026-09-29

  FineWeb paper HTML lists Dolma among the compared corpora: "the Common Crawl-based part of Dolma 1.6 (3T tokens) and 1.7 (1.2T tokens)", and states that the FineWeb model "outperforms all other open datasets we compared on our aggregate group of tasks".