Free🇺🇸
Multilingual pretraining dataset from Hugging Face covering more than 1,000 languages, including a large German subset.
FineWeb-Edu overview and key facts (October 2026)
Hugging Face pretraining dataset of 1.3 trillion tokens of educational web pages, selected from Common Crawl by an AI quality classifier.
Free
What is FineWeb-Edu?
FineWeb-Edu is an open pretraining dataset with 1.3 trillion tokens of English web pages rated as educational. Teams use it to train or continue training language models on higher-quality web text.
The dataset is a filtered subset of FineWeb and covers Common Crawl dumps since 2013; version 1.4.0 from July 2025 added snapshots up to June 2025. A larger variant with 5.4 trillion tokens uses a lower threshold. Sample configurations with 10, 100 and 350 billion tokens allow smaller experiments. Hugging Face maintains the dataset and published the method in a paper on arXiv.
FineWeb-Edu AI features and use cases
- Trains small and midsize modelsTeams training their own compact language models use the dataset as the main web corpus. The educational filter removes much low-value text, which helps smaller models learn more from the same compute budget.
- Reuses the quality classifierThe classifier that rated each page is published with its code. Data teams can apply it to their own crawls or internal documents to keep only instructive text before training.
- Runs quick ablationsThe 10 and 100 billion token samples let researchers compare data mixes or training settings on a small scale before committing to a full run. Results from the sample runs help estimate what a full training run would bring.
What sets FineWeb-Edu apart
- Selection by an AI classifier whose model and code are public, so the filtering is reproducible.
- Previous versions remain available in separate branches of the repository.
Where FineWeb-Edu is strongest
United States Available internationally
FineWeb-Edu is an English-language dataset downloadable worldwide from the Hugging Face Hub.
Questions to ask before you buy FineWeb-Edu
The data comes from public web pages and can contain personal data and copyrighted text; ODC-By requires attribution and the underlying Common Crawl terms also apply. Plan storage and compute for terabyte-scale downloads, or start with a sample configuration. Document the dataset in your model documentation if you fall under EU AI Act transparency duties for general-purpose models.
FineWeb-Edu reviews
No reviews yet. Be the first to share how your team uses it.
Every review shows its verification level. We read each one before it goes live. How we check reviews
Questions about FineWeb-Edu
Can we use FineWeb-Edu commercially?
The dataset is released under ODC-By 1.0, which permits commercial use with attribution. The terms of use of Common Crawl, the source of the pages, also apply.
How large is it?
The main version holds 1.3 trillion tokens; a variant with a lower threshold holds 5.4 trillion. Samples of 10, 100 and 350 billion tokens exist for smaller projects.
Is it only English?
Yes, FineWeb-Edu is English. For other languages, including German, Hugging Face maintains the multilingual FineWeb2.
FineWeb-Edu update historyno updates yet
No changes recorded for FineWeb-Edu yet.
Changes the vendor made and confirmed user reports, each checked against the vendor's own pages. We review every tool monthly.
Alternatives to FineWeb-Edu
Tools in the same category (Training data) with the most overlapping markets.
Free🇺🇸
Instruction-tuning mix of 939,344 examples from the Allen Institute for AI, used to train the Tulu 3 models.
Free🇫🇷
Open pretraining corpus of 2.27 trillion tokens from Pleias in Paris, containing only uncopyrighted or freely licensed documents.
Contact for pricing · Free planAvailable internationally
Free, open-source prompt library, formerly Awesome ChatGPT Prompts, available as a website, MCP server or self-hosted company library.