AI Services Directory

Embedding / RAG · Updated 2026-10-25

OpenAI Embeddings — what it is actually for

OpenAI's embedding models convert text into vectors for semantic search, clustering, and retrieval-augmented generation. Embeddings bill input tokens only and are priced per 1M tokens.

Short answer: use OpenAI Embeddings when you need semantic search over documents. It is a usage-based api service from OpenAI; the official pricing source is linked below.

Vendor
OpenAI
Category
Embedding / RAG
Pricing model
Usage-based API
Free option
No separate free API tier is published.

What OpenAI Embeddings is used for

  • Semantic search over documents
  • RAG retrieval pipelines
  • Clustering and classification

Source: official product page and official documentation.

Who OpenAI Embeddings is for

  • Developers already using OpenAI models
  • Simple embedding pipelines with predictable token costs
  • Apps that need dimension truncation

Source: official documentation.

Who it is not for

  • Teams that need domain-specific or multilingual-specialized embeddings
  • Users who need large-scale free embeddings

Source: documented plans, limits, and deployment model on the official product page.

Pricing

  • text-embedding-3-small is listed at US$0.02 per 1M input tokens; text-embedding-3-large is listed at US$0.13 per 1M input tokens.Source: openai.com
  • Batch API requests are billed at a 50% discount for embeddings.Source: platform.openai.com
  • Embeddings have no output tokens; only input tokens are billed.Source: platform.openai.com

Models support reduced output dimensions, which can lower vector storage costs. Check the current pricing page for rates.

Full pricing page: https://openai.com/api/pricing/

Alternatives to OpenAI Embeddings

  • Cohere Embed — Embed 5 Pro and Fast models with separate indexing and query pricing.
  • Voyage AI — Embedding and reranking models for RAG, with domain-specific variants.

Official sources

OpenAI Embeddings FAQ

Are OpenAI embeddings free?

No separate free API tier is published. Embeddings are billed per input token.

How much do OpenAI embeddings cost?

The official pricing page lists text-embedding-3-small at US$0.02 and text-embedding-3-large at US$0.13 per 1M input tokens.

Do embeddings have output token costs?

No. Embedding requests are billed on input tokens only.

See also: compare all services by use case · about this directory.