Multilingual-E5

Multilingual-E5: a state-of-the-art text embedding model supporting 100+ languages.

Multilingual-E5, developed by Intfloat, delivers high-quality text embeddings across over 100 languages. This model excels in multilingual information retrieval and semantic search tasks, outperforming previous benchmarks. For instance, it achieves strong results on the MIRACL benchmark, demonstrating its cross-lingual capabilities. Its 'large' variant (multilingual-e5-large) is particularly effective for diverse applications, from clustering documents in different languages to powering multilingual question-answering systems. The model is readily available on Hugging Face, facilitating easy integration and deployment for developers working with global text data.

https://huggingface.co/intfloat/multilingual-e5-large