Hugging Face Updates Sentence Transformers with Multimodal Embedding and Reranker Support

Hugging Face announced an update to its Sentence Transformers Python library (v5.4), which now supports multimodal embedding and reranker models. This update allows users to encode and compare texts, images, audio, and videos within a shared embedding space, facilitating cross-modal search, visual document retrieval, and multimodal RAG pipelines. The library provides capabilities for both embedding models, which map inputs into a common space, and reranker models, which score the relevance of mixed-modality pairs.