Overview

Released in January 2024 together with text-embedding-3-small, this model converts text into numerical vectors that capture meaning, the basis of semantic search and retrieval-augmented generation. Its embeddings can be shortened to fewer dimensions to save storage with modest quality loss.

Highlights

  • Embeddings up to 3,072 dimensions
  • Dimensions can be reduced to trade quality for storage
  • Input up to 8,191 tokens
  • Improved multilingual retrieval over earlier OpenAI embeddings
  • Low price per token compared with chat models

Typical uses

Semantic search, RAG pipelines, recommendation, clustering and deduplication of text.

Things to consider

  • Embeddings from different models are not compatible; switching requires re-embedding all documents.
  • Data is processed by OpenAI’s API.
  • Benchmark leaders change often; compare on your own data.

Specifications are taken from the provider’s official announcement, documentation or model card. Benchmark results depend on evaluation settings; see how AI benchmarks work.