BTC
ETH
SOL
XRP
BNB
TON
DOT
LTC
POL
HBAR
NEAR
ARB
UNI
A list of numbers that captures the meaning of a piece of text or an image, so a computer can find things that mean similar things.
AI Drafted with AI help · Reviewed by Keshav Aggarwal · Updated
In simple words
Imagine a giant map where every word has its own spot. “Dog” sits near “puppy”, and far from “spaceship”.
An embedding is a word’s address on that map.
The technical side
An embedding model turns input into a vector, often hundreds or thousands of numbers long. Inputs with similar meaning end up close together, measured by cosine similarity or distance. Embeddings power semantic search, recommendations, clustering and the retrieval step in RAG.
Embedding models are much cheaper to run than chat models and are priced per million input tokens.
Embedding kya hai?
Embedding text ka matlab numbers mein badal deta hai. Milte-julte matlab wale shabd numbers mein bhi paas-paas aate hain.
Related terms
- Vector DatabaseAIA database built to store embeddings and quickly find the ones closest in meaning to a question.
- RAG (Retrieval-Augmented Generation)AIA way to make an AI answer from trusted documents: it first looks up the relevant pages, then writes its answer from them.
- LLM (Large Language Model)AIAn AI model trained on huge amounts of text so it can read and write language, like the models behind ChatGPT, Claude and Gemini.