Generate Text Embeddings with OpenAI

Text embeddings convert natural language into dense numeric vectors that capture semantic meaning. Two sentences with similar meaning produce similar vectors (high cosine similarity) even if they share no words, while semantically unrelated sentences produce distant vectors. OpenAI's text-embedding-3-small and text-embedding-3-large models are the standard choice for semantic search, RAG pipelines, clustering, and classification tasks that require meaning-aware text representations. This example shows an embeddings API request for a set of sentences and explains how to interpret the resulting vectors for similarity search. The text-embedding-3-small model produces 1,536-dimensional vectors and costs $0.02 per million tokens — about 2,000 times cheaper than GPT-4o for the same text volume. For most RAG applications, text-embedding-3-small is the right default; use text-embedding-3-large only when you have benchmarked a quality gap on your specific data. For RAG pipelines, embed both your knowledge base documents (at indexing time) and the user's query (at search time), then find the top-k documents with highest cosine similarity. The dimensionality parameter allows you to reduce vector dimensions for storage efficiency — reducing from 1,536 to 512 dimensions cuts storage and search latency by 3x with only modest quality reduction.

{
  "model": "text-embedding-3-small",
  "input": [
    "How do I reset my password?",
    "What is the process for changing my account password?",
    "I forgot my login credentials",
    "How do I cancel my subscription?"
  ],
  "dimensions": 1536,
  "encoding_format": "float"
}