Embeddings
Dense numerical vector representations of text (or images, code, audio) where semantically similar inputs map to nearby vectors.
An embedding model maps a token sequence to a fixed-length vector (commonly 768 to 3072 dimensions). Cosine similarity in that space approximates semantic similarity. Most 2026 production embedding pipelines use OpenAI's text-embedding-3-large, Voyage-3, or open-source bge-large. Embeddings power RAG, deduplication, clustering, classification, and recommendation.
Related terms
An AI pattern that retrieves relevant documents from a vector database and injects them into the LLM prompt — so the model can answer from custom knowledge it was not trained on.
A database optimized for similarity search over high-dimensional embedding vectors — the backbone of RAG and semantic search.
A neural network with billions of parameters trained on broad text corpora to predict and generate language — the engine behind ChatGPT, Claude, and Gemini.
Need this built into a real product?
Viprasol Tech ships production code for everything defined here — MT4/MT5 EAs, AI agents, B2B SaaS, AWS architecture.
Send a brief →