Definition
Full definition of embedding
You pass text to an embedding model (OpenAI's text-embedding-3, Cohere's embed-v3, Voyage's voyage-3) and get back a list of floats (typically 1024-3072 dimensions). Two pieces of text with similar meaning produce similar vectors, so you can search "by meaning" instead of "by keyword". The foundation of RAG, semantic search, recommendation, and clustering.
In practice
Embedding examples
Embedding output
Input: 'customer cancelled'. Output: [0.012, -0.089, 0.234, ... ] (1536 floats)
Used by
Apps that exemplify embedding
See embedding in action across real integrations.
FAQ
Common questions about embedding
Which embedding model is best?
For most tasks: OpenAI's text-embedding-3-small (cheap) or text-embedding-3-large. Cohere and Voyage often outperform on specific domains; benchmark on your data.
Why are embeddings expensive at scale?
They're cheap to generate (~$0.02 per 1M tokens), but storage and search costs scale linearly with corpus size. Plan for ~$0.50/month per million docs.
