This dataset contains 4.17 billion tokens of the highest quality educational content, carefully filtered from the FineWeb-Edu dataset's 2025 Common Crawl snapshots. This represents the cream of the crop - only the top ~2% of documents that meet strict quality criteria.
