Microsoft Research developed the E5 text embedding model family, maintained on Hugging Face under the intfloat account.

E5 is a family of text embedding models, including e5-large-v2 and multilingual-e5-large, trained via weakly-supervised contrastive pretraining for retrieval and semantic search.

Weights are hosted on Hugging Face under the intfloat organization, with code in the microsoft/unilm GitHub repository.

Original E5 released in December 2022, with multilingual-e5 variants following in 2023.

Built to provide strong general-purpose text embeddings competitive with proprietary APIs for search and RAG systems.

Used via Sentence-Transformers or Hugging Face Transformers.

MIT licensed, fully open weights, no credentials required, self-hosted; maintained by Microsoft (code) and the intfloat community account (weights).