Home / Guides / OpenAI embedding pricing: text-embedding-3 planning

OpenAI embedding pricing: text-embedding-3 planning

How OpenAI embedding API pricing works for planning: token × $/1M, text-embedding-3-small vs large tradeoffs, Batch vs standard, without inventing live unit prices.

OpenAI embedding cost in one formula

OpenAI bills embeddings by input tokens of the strings you embed. Cost ≈ (tokens / 1,000,000) × price per million for that embedding model. There is no chat-style output token line for a normal embed call.

Dimensions you request mainly change retrieval quality and vector storage size. Always verify current unit prices on OpenAI’s pricing page. Rates change; TokenCalculator presets are editable reference defaults, not catalog-verified live prices.

text-embedding-3-small vs text-embedding-3-large

Large is usually priced higher per million tokens than small. Paste both live rates; do not invent a multiplier. Large often wins on hard retrieval evals; small often wins on cost and latency for large corpora. Higher default dimensions can improve recall and raise vector DB storage.

Changing model or dimensions usually means a full re-embed. Budget that as a reindex event. Legacy text-embedding-ada-002 still appears in older stacks: price it from the current page and plan migration when you reindex.

Batch vs standard for OpenAI embeds

OpenAI’s Batch API often discounts eligible offline jobs, including embeddings when offered for that model. Use Batch for initial index builds and scheduled full reindexes when you can wait. Keep interactive query embedding on the standard endpoint.

Toggle batch in the embedding cost calculator and paste the batch rate from the pricing page when it differs.

OpenAI embed cost vs generation cost in RAG

If you use OpenAI for both embeddings and chat, separate the meters. Embed covers corpus build/refresh plus optional query embeds. Generate covers system + question + retrieved chunks + output at chat model rates.

Generation usually dominates busy RAG apps because retrieved context re-bills every query. Model both: embedding cost calculator for the index side, RAG cost calculator for the full pipeline with catalog generation rates.

Common mistakes

Avoid these OpenAI embedding pricing mistakes.

  • Quoting a blog’s $/1M as if it were still live
  • Comparing small vs large on price without a retrieval eval
  • Forgetting Batch on a multi-hour offline index
  • Changing dimensions or model without budgeting a full re-embed
  • Counting unique document tokens while the API sees overlapped chunks
  • Blending OpenAI chat spend and embed spend into one opaque number

Frequently asked questions

How is OpenAI embedding API pricing calculated?
Tokens of text sent to the embeddings endpoint ÷ 1,000,000 × the published $/1M for that model (and Batch vs standard if applicable). Verify on OpenAI’s pricing page.
How much does text-embedding-3-small cost?
Whatever OpenAI currently lists. Paste that number into /tools/embedding-cost. Do not use evergreen articles as a price table.
How much does text-embedding-3-large cost?
Copy from OpenAI’s pricing page, paste into the calculator, and compare against small on your corpus tokens and evals.
Do embeddings charge for output tokens?
Generally no for standard embedding calls. You pay for input tokens of the text being embedded. Confirm in OpenAI’s current docs.
Does the Batch API reduce embedding cost?
When OpenAI offers Batch pricing for that embedding model and your job is eligible, yes, often at a lower $/1M with higher latency.
Are embedding dimensions billed separately on OpenAI?
Usually not as a separate API line. You still pay token rates; storage cost may change with vector size.
Small vs large: is a big price multiple worth it?
Run retrieval evals on your data. Then price index and refresh with both live rates in the calculator. Quality first, then cost.
Is embedding cost or generation cost higher in RAG?
Often generation for busy apps. Compare with /tools/rag-cost using catalog chat rates and your pasted embed rate.
Is text-embedding-ada-002 still priced?
If it remains listed, price it from the live page. Plan migration cost as a full reindex when you move to text-embedding-3.
Where is the free OpenAI embeddings pricing calculator?
TokenCalculator’s /tools/embedding-cost is free to use. You supply the live $/1M from OpenAI.

Try it in TokenCalculator

Pick a text-embedding-3 preset, replace the reference $/1M with the number from OpenAI’s pricing page, then enter corpus tokens and refresh cadence.

Open embedding cost calculator · Open cost calculator · Estimate API spend · All guides

Related tools

Embedding cost calculator · RAG cost · Batch pricing

Related guides

Embedding cost calculator: estimate corpus spend · Estimate embedding tokens · Batch embedding cost savings · Embedding reindex cost: how often to re-embed · Embedding vs generation cost in RAG

Sources and references

Official documentation used for definitions, counting methods, or rate cards. Always confirm critical budgets on the provider page.