Vibeleaderboard
Intfloat

model publisher

Intfloat

This is a Model Atlas entity, not a ranked standalone lab. A researcher-led model namespace rather than a separately documented organization.

E5 helped normalize a general-purpose embedding recipe that works across retrieval and semantic tasks without building a separate architecture for each application.

Models in the Atlas →

Profile

Overview

Intfloat is the Hugging Face publisher behind the E5 family of text embeddings. E5 models use a shared text-to-text contrastive framework for semantic similarity, retrieval, clustering, and related representation tasks.

The line includes English and multilingual encoders at several sizes. Public identity and organizational information are limited, while the models themselves are documented through papers, model cards, and broad downstream adoption.

Notable contributions

  1. 01Text-to-text contrastive embeddingsE5 frames diverse embedding tasks as text-pair learning and transfers that representation across retrieval and classification settings.