FAISS — Facebook AI Similarity Search — is an open-source library from Meta for efficient similarity search over dense vectors. It is not a full database but a highly optimised toolkit of index structures and search algorithms that developers embed directly in their applications.
FAISS has been enormously influential, providing fast, battle-tested implementations of many ANN approaches including flat (exact) indexes, IVF clustering, product quantisation, and HNSW graphs. It supports both CPU and GPU execution and can handle billions of vectors, which made it a default building block for research and early production vector search before purpose-built databases matured.
Its strength is raw performance and flexibility; its limitation is that it is a library, not a managed system. FAISS does not provide persistence, metadata filtering, replication, an API server, or operational tooling out of the box — you build those around it. Many full vector databases use FAISS or FAISS-like techniques internally while adding the storage, filtering, and infrastructure layers that production systems need.