FineWeb
by Hugging Face · open-source · Last verified 2026-03-17
A 15 trillion token, high-quality English web dataset curated by Hugging Face from 96 CommonCrawl snapshots using a carefully designed multi-stage filtering pipeline. FineWeb achieves state-of-the-art pretraining performance among open datasets and ships with FineWeb-Edu, a 1.3T token educational content subset filtered using an LLM-based quality classifier.
https://huggingface.co/datasets/HuggingFaceFW/fineweb ↗B+
B+—Good
Adoption: AQuality: A+Freshness: A+Citations: B+Engagement: F
Specifications
- License
- ODC-BY
- Pricing
- open-source
- Capabilities
- language-modeling, pretraining, educational-content
- Integrations
- hugging-face, datatrove
- Use Cases
- llm-pretraining, research, educational-model-training
- API Available
- Yes
- Tags
- nlp, pretraining, web-crawl, quality-filtered, hugging-face
- Added
- 2026-03-17
- Completeness
- 100%
Index Score
70.1Adoption
80
Quality
93
Freshness
92
Citations
78
Engagement
0
Put AI to work for your business
Deploy this dataset alongside autonomous AaaS agents that handle tasks end-to-end — no babysitting required.