Xet Dedup Saves Hugging Face 80% of Storage Globally, Engineers Say

lhoestq · x · 2026-10-22

Detailing the storage behind the new hf:// DuckDB path: Hugging Face Buckets are built on Xet, and global deduplication saves 80% of storage. The same file takes less space on HF than on any other filesystem or cloud provider — a key reason the hf:// workflow is efficient.

Related event: Hugging Face Ships DuckDB 2.0 Extension for SQL Access to Datasets(3 posts)→

Original post →

More from Infra

Infra channel →