Embedding API pricing
10 embedding models · USD per 1M tokens
| Model | Provider | $/1M ↑ | Dims | Ctx | MTEB | |
|---|---|---|---|---|---|---|
| OpenAI | $0.02 | 1536 | 8K | 62.3 | docs ↗ | |
| Voyage AI | $0.02 | 1024 | 32K | 64.7 | docs ↗ | |
| Jina AI | $0.02 | 1024 | 8K | 58.0 | docs ↗ | |
 Titan Text Embeddings V2 | Amazon | $0.02 | 1024 | 8K | 55.0 | docs ↗ |
| Voyage AI | $0.06 | 1024 | 32K | 66.4 | docs ↗ | |
| Mistral | $0.10 | 1024 | 8K | 59.0 | docs ↗ | |
| Nomic | $0.10 | 768 | 8K | 53.0 | docs ↗ | |
| Cohere | $0.12 | 1536 | 128K | 66.0 | docs ↗ | |
| OpenAI | $0.13 | 3072 | 8K | 64.6 | docs ↗ | |
| $0.15 | 3072 | 2K | 68.3 | docs ↗ |
compare.hint · Price per 1M input tokens. MTEB = retrieval score (approx). Dimensions shown are the model's maximum.
Reading this comparison
- 10 embedding models tracked, from $0.020 to $0.15 per 1M tokens - a 8x range.
- Embedding a 10M-token corpus costs roughly $0.20 to $1.50 depending on the model - a one-time or per-ingest cost, unlike the ongoing cost of storing and querying the resulting vectors.
- Vector dimensions range from 768 to 3072 across these models - more dimensions mean more vector storage downstream (see the vector database comparison).
- Among the 10 models that publish an MTEB retrieval score, the average is 61.7 - a higher score suggests better semantic search quality, but is not necessarily tied to price.
This price covers generating the vectors only. Storing and querying them is billed separately by whichever vector database you choose.