Embedding API pricing
Prices per 1M input tokens for 25 embedding models. "10K pages" assumes about 5M tokens (10,000 pages × ~400 words). Smaller vectors (fewer dimensions) also cut vector database storage costs.
| Model | Provider | Price /1M | Dimensions | 10K pages |
|---|---|---|---|---|
| text-embedding-3-small | OpenAI | $0.02 | 1536 | $0.10 |
| voyage-3-lite | Voyage AI | $0.02 | — | $0.10 |
| voyage-3.5-lite | Voyage AI | $0.02 | — | $0.10 |
| voyage-4-lite | Voyage AI | $0.02 | 1024 | $0.10 |
| voyage-3 | Voyage AI | $0.06 | — | $0.30 |
| voyage-3.5 | Voyage AI | $0.06 | — | $0.30 |
| voyage-4 | Voyage AI | $0.06 | 1024 | $0.30 |
| embed-v5.0-fast | Cohere | $0.08 | 2048 | $0.40 |
| embed-english-light-v3.0 | Cohere | $0.10 | — | $0.50 |
| embed-english-v3.0 | Cohere | $0.10 | — | $0.50 |
| embed-multilingual-v3.0 | Cohere | $0.10 | — | $0.50 |
| mistral-embed | Mistral | $0.10 | — | $0.50 |
| embed-v4.0 | Cohere | $0.12 | 1536 | $0.60 |
| embed-v5.0-pro | Cohere | $0.12 | 2048 | $0.60 |
| voyage-4-large | Voyage AI | $0.12 | 1024 | $0.60 |
| voyage-code-4 | Voyage AI | $0.12 | 1024 | $0.60 |
| voyage-context-4 | Voyage AI | $0.12 | 1024 | $0.60 |
| voyage-multimodal-3 | Voyage AI | $0.12 | — | $0.60 |
| voyage-multimodal-3.5 | Voyage AI | $0.12 | 1024 | $0.60 |
| text-embedding-3-large | OpenAI | $0.13 | 3072 | $0.65 |
| gemini-embedding-001 | $0.15 | 3072 | $0.75 | |
| codestral-embed | Mistral | $0.15 | — | $0.75 |
| voyage-3-large | Voyage AI | $0.18 | — | $0.90 |
| voyage-code-3 | Voyage AI | $0.18 | — | $0.90 |
| voyage-context-3 | Voyage AI | $0.18 | — | $0.90 |
FAQ
How many tokens is my document set?
Roughly 1,300 tokens per 1,000 English words. 10,000 pages of about 400 words each is around 5M tokens. Use the token counter to check a sample.
Is embedding the expensive part of RAG?
Rarely. Embedding is a one-time cost per document (plus updates). The generation model reading retrieved chunks on every query usually dominates the monthly bill.