Loading the catalogue…
BAAI's open-weight multilingual text-embedding model for retrieval, similarity search, classification and clustering.
| Inference Provider | Sovereignty | Runs in | Price | Status | Pin |
|---|---|---|---|---|---|
| L3 | €0.10 per 1M input tokens | Online | |||
| L0 | €0.07 per 1M input tokens | Online |
11 use cases classified by the curator. The note says why.
Purpose-built multilingual embedding model with an 8192-token context window, suited to dense retrieval over long passages across languages.
Embedding outputs can feed lightweight classifiers or similarity-based labelling, but it has no native generative classification head, so it is a competent rather than primary choice.
Embedding-only model with no text-generation capability, so it cannot produce answers from retrieved documents.
Useful only as the retrieval stage of a document Q&A pipeline; it has no generative capability to produce answers on its own.
Entity extraction needs generative or token-labelling output; this embedding model does not produce extracted spans or structured fields.
No function-calling support is documented, and the model produces vectors rather than structured tool invocations.
Text-only embedding model with no vision input capability.
No text-generation capability is documented, so it cannot draft long-form content.
Sourced numbers, not our own evaluations. Percentiles are against the embedding models on Stav.
curl https://api.stav.ai/v1/embeddings \
-H "Authorization: Bearer $STAV_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "baai/bge-multilingual-gemma2",
"input": ["Hello, world"]
}'→ X-Stav-Request-Id · X-Stav-Model · X-Stav-Provider → X-Stav-Sovereignty-Level — the rung the request was served at