Embeddings cost calculator
Compare embedding costs across OpenAI, Cohere, Voyage, and Mistral. Adjust corpus size, re-embed frequency.
Embedding costs add up fast on large corpora. Here’s what it actually costs to embed your corpus across the major providers.
| Model | Dim | $ / 1M tokens | One-time cost | Upfront | vs cheapest |
|---|---|---|---|---|---|
text-embedding-3-small OpenAI | 1,536 | $0.02 | $0.07 | $0.07 | cheapest |
voyage-3-lite Voyage AI | 512 | $0.02 | $0.07 | $0.07 | — |
voyage-3 Voyage AI | 1,024 | $0.06 | $0.22 | $0.22 | 3.00× |
Embed v3 English Cohere | 1,024 | $0.10 | $0.38 | $0.38 | 5.00× |
Embed v3 Multilingual Cohere | 1,024 | $0.10 | $0.38 | $0.38 | 5.00× |
mistral-embed Mistral | 1,024 | $0.10 | $0.38 | $0.38 | 5.00× |
text-embedding-3-large OpenAI | 3,072 | $0.13 | $0.49 | $0.49 | 6.50× |
Token count uses the ceil(chars / 4) heuristic per document — accurate within ±10% for English prose. Non-Latin scripts and code tokenize denser; expect 10–25% more tokens than shown. Larger embedding dimensions are not proportionally more expensive — providers price per input token, not per output vector.
One-time cost = total tokens × price per million. Monthly cost multiplies the one-time pass by the re-embed frequency (weekly ≈ 4.33 passes/month, daily ≈ 30.4 passes/month). Storage and query costs (Pinecone, pgvector, Qdrant) are not included.
Source: provider pricing pages, May 2026. Token count estimated as chars / 4.
See per-embedding cost on real traffic — try Tokenwise.