Cheapest Embedding Models
Ranked from provider-published pricing · Prices checked 12 August 2026
Embedding models turn text into vectors for search, clustering and retrieval. They only consume input tokens, so this list ranks purely on published input price per million tokens.
Embedding cost is usually a small share of a RAG system’s bill compared with the generation model, but it scales with corpus size, and re-embedding a large corpus after a model change can be a significant one-off.
The ranking
top 15 of 15text-embedding-3-small from OpenAI leads this ranking at $0.020/M. That is 80% below the $0.100/M median across the 15 models that qualify for this list.
| # | Model | Provider | Input Price per 1M Tokens | Context |
|---|---|---|---|---|
| 1 | text-embedding-3-small | OpenAI | $0.020/M | 8,000 |
| 2 | Titan Embeddings V2 | Amazon | $0.020/M | 8,000 |
| 3 | text-embedding-004 | $0.025/M | 2,000 | |
| 4 | voyage-3-lite | Voyage AI | $0.065/M | 32,000 |
| 5 | mistral-embed | Mistral | $0.100/M | 8,000 |
| 6 | Embed v3 English | Cohere | $0.100/M | 500 |
| 7 | Embed v3 Multilingual | Cohere | $0.100/M | 500 |
| 8 | Solar Embedding | Upstage | $0.100/M | 4,000 |
| 9 | voyage-3 | Voyage AI | $0.120/M | 32,000 |
| 10 | voyage-finance-2 | Voyage AI | $0.120/M | 32,000 |
| 11 | voyage-law-2 | Voyage AI | $0.120/M | 32,000 |
| 12 | voyage-multilingual-2 | Voyage AI | $0.120/M | 32,000 |
| 13 | text-embedding-3-large | OpenAI | $0.130/M | 8,000 |
| 14 | voyage-3-large | Voyage AI | $0.180/M | 32,000 |
| 15 | voyage-code-3 | Voyage AI | $0.180/M | 32,000 |
How this list is built
- Ranked cheapest first by published input price per million tokens.
- Limited to embed models.
- Includes models marked available or preview; retired and deprecated SKUs are excluded.
- Prices are the rates each provider publishes, not estimates. How the blended rate is calculated.
Frequently asked questions
How much does it cost to embed a large document set?
Estimate the total tokens in the corpus and multiply by the model’s input rate. A million tokens (roughly 750,000 words) costs the model’s per-million input price once. The recurring cost is usually much smaller, since you only embed new or changed documents plus each incoming query.
Related rankings
Price it for your workload
Rankings are recomputed on every deploy from the catalogue. Prices checked 12 August 2026.