AI & LLM
Token, cost & context tools for LLMs
OpenAI Embedding API Cost — text-embedding-3-small Pricing
Embedding models convert text into vector representations — there is no generated output, only input tokens. OpenAI's text-embedding-3-small is priced at $0.02/M input tokens, making it exceptionally cheap for retrieval-augmented generation (RAG) and semantic search. At 1,000 input tokens per document chunk, each embedding call costs $0.00002: $0.02 per 1,000 chunks, $20.00 to embed 1 million chunks. Embedding costs are typically a one-time indexing cost; retrieval queries are far shorter and even cheaper.
text-embedding-3-small: $0.02/M input tokens (no output tokens — embeddings only)
- Cost per 1,000-token chunk: $0.00002
- Per 1,000 chunks: $0.02
- Per 1,000,000 chunks: $20.00
- Context window (max chunk size): 8,192 tokens
- Rates as of 2026-07-03
Frequently asked questions
How much does the OpenAI Embeddings API cost?
text-embedding-3-small costs $0.02/M input tokens — at 1000 tokens per chunk that is $0.00002/chunk and $0.02/1,000 chunks. Embedding a 1,000-document knowledge base (1,000 tokens each) costs $0.02. Embedding models have no output tokens and no per-call overhead charge. Rates as of 2026-07-03.
What is the maximum chunk size for text-embedding-3-small?
text-embedding-3-small accepts up to 8,192 tokens per API call. For RAG systems, common chunk sizes range from 256 to 512 tokens to balance retrieval precision and context completeness. Larger chunks are cheaper per chunk but may reduce retrieval quality. The model produces 1,536-dimensional vectors by default (configurable down to lower dimensions for storage savings).