Comment by ignoramous

Mistral has published large language models, not embedding models? sgrep uses Google's Word2Vec to generate embeddings of the corpus and perform similarity searches on it, given a user query.