BGE-M3
BGE-M3 is BAAI’s open-weight multilingual embedding model, covering more than 100 languages. It turns inputs of up to 8,192 tokens into 1,024-dimensional vectors for semantic search, retrieval and clustering across languages.
Quickstart
OpenAI-compatible: point any OpenAI SDK at the Venice base URL with a Venice API key.
curl https://api.venice.ai/api/v1/embeddings \
-H "Authorization: Bearer $VENICE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "text-embedding-bge-m3",
"input": "Private AI for everyone."
}'https://api.venice.ai/api/v1Parameters
| Parameter | Values |
|---|---|
model | text-embedding-bge-m3 |
input | A string, or up to 2,048 strings |
encoding_format | float | base64 |
LangChain’s OpenAIEmbeddings sends token IDs, which Venice rejects. Set check_embedding_ctx_length=False.
Vectors
Each input comes back as 1,024 numbers. Stored as float32, each number takes 4 bytes.
Each square is 64 dimensions, and each row 1,024.
Similarities from different models aren’t comparable, so store the model ID with each vector.
Pricing
Billed per 1M input tokens, in USD.
Estimate a corpus
Specs
- Model ID
text-embedding-bge-m3- Endpoint
POST /api/v1/embeddings- Dimensions
- 1,024
- Max input
- 8,192 tokens per input
- Batch
- Up to 2,048 inputs per request
- Encodings
floatorbase64- Input
- Text only
- Privacy
- Private
- Weights
- BAAI/bge-m3
- Added
- Mar 14, 2025
Compare
Every embedding model on Venice, cheapest first.
| Model | Dimensions | Per 1M tokens |
|---|---|---|
| BGE-EN-ICL | 4,096 | $0.0125 |
| Multilingual E5 Large Instruct | 1,024 | $0.0125 |
| Nemotron Embed VL 1B v2 | 2,048 | $0.0125 |
| Qwen3 Embedding 0.6B | 1,024 | $0.0125 |
| Qwen3 Embedding 8B | 4,096 | $0.0125 |
| Text Embedding 3 Small | 1,536 | $0.025 |
| BGE-M3 | 1,024 | $0.15 |
| Text Embedding 3 Large | 3,072 | $0.1625 |
| Gemini Embedding 2 Preview | 3,072 | $0.25 |