Sentence Transformers releases 400x faster static embedding models: static-retrieval-mrl-en-v1 and static-similarity-mrl-multilingual-v1

AI Impact Summary

The blog describes training strategies that yield static embedding models 100x-400x faster on CPU, enabling on-device and edge deployments for English retrieval and multilingual similarity tasks. By leveraging contrastive learning and Matryoshka Representation Learning (MRL), these models aim to preserve quality while dramatically reducing inference time and resource usage. The two released models, sentence-transformers/static-retrieval-mrl-en-v1 and sentence-transformers/static-similarity-mrl-multilingual-v1, can be used with the Sentence Transformers library on CPU, enabling familiar workflows without GPU infrastructure.