BGE-M3 pricing
BGE-M3 from BAAI — embeddings API at $0.010 per 1M tokens, 1024 dimensions, 8,192-token max input. Open weights — self-host for free; multilingual, dense + sparse + multi-vector.
BAAI/bge-m3open weights Billed per input token · prices USD, last verified 2026-06-27.
BGE-M3 cost calculator
Estimate the bill for embedding a corpus. Enter total tokens (≈ words × 1.3).
Re-embedding on model changes repeats this cost. Compare all models on the embeddings pricing page.
BGE-M3 vs other embedding models
Cost per 1M tokens, cheapest first — BGE-M3 highlighted.
| Model | Provider | Per 1M | Dimensions | Max input |
|---|---|---|---|---|
| BGE-M3open | BAAI | $0.010 | 1024 | 8K |
| text-embedding-3-small | OpenAI | $0.020 | 1536 | 8K |
| Jina Embeddings v3 | Jina AI | $0.020 | 1024 | 8K |
| Voyage-3.5 | Voyage AI | $0.060 | 1024 | 32K |
| Mistral Embed | Mistral | $0.100 | 1024 | 8K |
| Cohere Embed v4 | Cohere | $0.120 | 1536 | 128K |
| text-embedding-3-large | OpenAI | $0.130 | 3072 | 8K |
| Gemini Embedding | $0.150 | 3072 | 2K |
How BGE-M3 pricing works
BGE-M3 bills per input token at $0.010 per million — there are no output tokens, since the response is a fixed 1024-dimension vector. Cost scales only with how much text you embed, so the levers are corpus size and how often you re-embed. Because it is open weights, you can also self-host and pay only your own compute.
Higher dimensions improve retrieval marginally but increase vector-database storage and query cost proportionally — many models (including this one where noted) support shortening dimensions to trade a little accuracy for big storage savings.
BGE-M3 pricing FAQ
How much does BGE-M3 cost?
BGE-M3 costs $0.010 per 1M input tokens. Embedding 1 million typical documents of ~500 tokens each (500M tokens) would cost about $5.00.
How many dimensions does BGE-M3 output?
BGE-M3 returns 1024-dimension vectors. Its max input is 8,192 tokens per request.
Is BGE-M3 a cheap embedding model?
BGE-M3 is among the cheapest embedding APIs tracked, at $0.010 per 1M tokens.
Is BGE-M3 open weights?
Yes — BGE-M3 is open weights, so you can self-host it for free instead of paying per-token API rates.