Embeddings are dense numeric vectors — typically 384 to 3,072 dimensions — that represent the semantic meaning of text, images, or other data. Two pieces of text with similar meaning map to nearby vectors, even if they share no words.
Embeddings are the foundation of semantic search, RAG systems, recommendation engines, deduplication, and clustering. Common providers: OpenAI text-embedding-3, Cohere Embed, Voyage AI, and open-source options like BGE and E5.
Once generated, embeddings are stored in a vector database (Pinecone, Weaviate, pgvector, Milvus) and searched using nearest-neighbour algorithms. Every serious LLM production system uses embeddings somewhere in the stack.