Hugging Face Introduces Storage Buckets for ML Artifacts
Action Required
Teams can now efficiently manage and transfer constantly changing ML artifacts, reducing storage costs and improving pipeline performance.
AI Impact Summary
Hugging Face is introducing Storage Buckets, a new object storage solution built on S3-like technology, designed for the dynamic nature of ML artifact creation. These buckets provide a mutable space for checkpoints, processed data, and other intermediate files, addressing the limitations of Git for managing constantly changing data. This capability is particularly beneficial for training pipelines and large-scale ML workloads, offering improved transfer speeds and more efficient storage through Xet's deduplication capabilities. Users will be able to sync data directly into buckets using the CLI or Python, and integrate with existing workflows via fsspec.
Affected Systems
- Date
- 10 Mar 2026
- Change type
- capability
- Severity
- high