Skip to main content
Back

Hugging Face Takes Over Sentence Transformers, Unifying 16,000 Open-Source Embedding Models

Sentence Transformers, the SBERT library behind over 16,000 embedding models on the Hub, moves from UKP Lab at TU Darmstadt to Hugging Face stewardship.

Model, Full Access
  • Sentence Transformers (SBERT) officially moves from TU Darmstadt's UKP Lab to Hugging Face stewardship.
  • Tom Aarsen, maintainer since late 2023, continues leading the project inside Hugging Face.
  • License stays Apache 2.0; project remains community-driven and open source.
  • Over 16,000 Sentence Transformers models on the Hub serve more than a million monthly unique users.
  • Repo moves to github.com/huggingface/sentence-transformers with CI and tighter transformers integration.
  • Recent roadmap: v3 modern training, v4 rerankers, v5 sparse encoders, plus multimodal embeddings.

Sentence Transformers, the open source framework behind many semantic search, retrieval, reranking, and clustering systems, is moving from the Ubiquitous Knowledge Processing Lab at TU Darmstadt to Hugging Face. Tom Aarsen, who has maintained the project since late 2023, will continue to lead development.

Nils Reimers developed the original Sentence-BERT work with UKP Lab director Iryna Gurevych in 2019. The resulting library, often called SBERT, made bi-encoders practical: these models convert text into vectors that applications can compare efficiently instead of running a language model over every possible pair.

The handoff in six lines

Area Current status Developer action
License Apache 2.0 None
Package sentence-transformers Keep the existing installation and import paths
API No transfer-specific migration announced Keep current code unless upgrading versions
Maintainer Tom Aarsen remains project lead None
Repository GitHub repository Update old UKPLab Git remotes
Documentation Project documentation Keep existing bookmarks

One home for the embedding stack

The announcement reports more than 16,000 public Sentence Transformers models on the Hugging Face Hub and over one million monthly unique users. Popular checkpoints include