Hugging Face Launches Buckets for Large-Scale AI Data
August 22, 2026
Hugging Face introduced Buckets, a new repository type powered by Xet technology for hosting extremely large files, such as 33TB datasets. Rerun has already utilized this to host the ABC-130k bi-manual teleoperation dataset.
HOW THIS AFFECTS YOU
●
builderYou can now host and share massive training datasets directly on Hugging Face with S3-like optimization.
●
researcherAccess to large-scale, specialized datasets like ABC-130k is becoming easier via centralized hosting.