all-MiniLM-L6-v2
sentence-transformers/all-MiniLM-L6-v2 · Apache-2.0
The 22M embedding model behind half the semantic search on the planet.
all-MiniLM-L6-v2 turns sentences into 384-dim vectors with a great quality/speed trade-off, the default choice for RAG, deduplication and clustering.
Score
Setup at a glance
pip install sentence-transformers Deep dive
all-MiniLM-L6-v2 is the 22M-parameter sentence-embedding model that became the default first choice for semantic search: it maps any sentence to a 384-dimensional vector that captures meaning, not keywords.
It is small enough to run on CPU at thousands of embeddings per second, which is why it underlies countless search, clustering and dedup pipelines — with over 200 million downloads on Hugging Face.
It is trained and distributed through the sentence-transformers library, so usage is a two-liner: load the model, encode your texts, and compare with cosine similarity.