The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Summary
The Pile is an 825 GiB English text corpus targeted at training large-scale language models, constructed from 22 diverse high-quality subsets, many of which derive from academic or professional sources. The authors report that models trained on the Pile improve significantly over both Raw CC and CC-100 on all Pile components. They document potentially concerning aspects of the data and make publicly available the code used in its construction. [S1]
Claims
- id: C1 claim: 'The Pile is an 825 GiB English text corpus for large-scale language model training.' source: S1 evidence: "an 825 GiB English text corpus targeted at training large-scale language models"
- id: C2 claim: 'It is constructed from 22 diverse high-quality subsets.' source: S1 evidence: "constructed from 22 diverse high-quality subsets"
- id: C3 claim: 'The authors make publicly available the code used in its construction.' source: S1 evidence: "We make publicly available the code used in its construction"
Verification
- Code available: Paper: https://github.com/EleutherAI/the-pile, which returned HTTP 200 on 2026-11-12. Not run.
- Reproduced: not reproduced (2026-11-12). Code not run.
- Related work: This batch also has a verification note for the later FineWeb datasets.
- Limits: The title says 800GB, while the abstract states 825 GiB; this note uses the 825 GiB figure (v1, 31 Dec 2020).
- License: CC BY 4.0 on the abs page (2026-11-12; icon by-4.0.png; http://creativecommons.org/licenses/by/4.0/). Submitted 31 Dec 2020 (v1). Sources: [S1] https://arxiv.org/abs/2101.00027, v1 (31 Dec 2020). checked_on: 2026-11-12.
When to apply
Use when tracing early open pretraining corpora, or the original 22-subset construction of the Pile.
Facts
- id: F1 claim: 'The Pile is an 825 GiB English text corpus constructed from 22 diverse high-quality subsets.' confidence: medium date: 2026-11-12 sources: [S1] license: CC BY 4.0
- id: F2 claim: 'The authors report that models trained on the Pile improve over Raw CC and CC-100 on all Pile components.' confidence: medium date: 2026-11-12 sources: [S1] license: CC BY 4.0
- id: F3 claim: 'They publicly released the code used to construct the dataset.' confidence: medium date: 2026-11-12 sources: [S1] license: CC BY 4.0
Sources
- [S1] arXiv 2101.00027, https://arxiv.org/abs/2101.00027. License: CC BY 4.0. v1 (31 Dec 2020). Accessed 2026-11-12.
License: this note is CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Summary by iinventedlight (AI-assisted). Paper is CC BY 4.0.
License: CC-BY-4.0 · Summary by iinventedlight (AI-assisted). Source: Leo Gao, Stella Biderman, Sid Black, and 9 coauthors (listed on the arXiv abs page), The Pile: An 800GB Dataset of Diverse Text for Language Modeling (CC BY 4.0).
Cite this
'iinventedlight (AI-assisted)'. 2026. "The Pile: An 800GB Dataset of Diverse Text for Language Modeling." iinventedlight Library, iil:note:1pmwr6kp, version of 2026-11-12. https://iinventedlight.com/library/notes/2101-00027-the-pile/. License: CC-BY-4.0.
@misc{iil_note_1pmwr6kp,
author = {{'iinventedlight (AI-assisted)'}},
title = {{The Pile: An 800GB Dataset of Diverse Text for Language Modeling}},
howpublished = {iinventedlight Library},
year = {2026},
date = {2026-11-12},
url = {https://iinventedlight.com/library/notes/2101-00027-the-pile/},
note = {Permanent ID iil:note:1pmwr6kp. License: CC-BY-4.0.},
}
Spotted a mistake? Use the MCP feedback tool or email hello@iinventedlight.com.
For language models and agents: start at /llms.txt | /agents.md | /.well-known/agent-library.json. This page in Markdown: /library/notes/2101-00027-the-pile.md.