AI Potluck
Model components / Training & synthetic datasets

FineWeb-Edu

Hugging Face

FineWeb-Edu is an educational subset of FineWeb (~1.3 trillion tokens) created by Hugging Face's FineWeb team. It was produced by filtering FineWeb with an educational-quality classifier trained on Llama-3-70B annotations, retaining text scored as highly educational. It is intended to improve LLM performance on knowledge- and reasoning-heavy benchmarks relative to unfiltered web data.

Verified live 2026-06-22 via primary sources. Permissive ODC-BY license with ungated download and full dataset card.

Openness

5 high confidence
5.0
license
ODC-BY
gated
false
dataset_card
present
format
parquet

Permissive ODC-BY license with ungated download and full dataset card.

Adoption

4 high confidence
4.0

384,910 monthly downloads on Hugging Face, graded on the training-corpus bands.

Capability

5 high confidence
5.0

Dominates FineWeb and all peers on MMLU/ARC, matching full FineWeb with ~10x fewer tokens; current default web corpus for SmolLM2/3 and Apertus.

Unchanged since 2026-07-04 (last edited, not re-checked)