There is no per-unit API price: self-host only.
There is no free API. The weights are open, so you can run it yourself.
#134 of 200 in Embeddings.
On MTEB Dutch v1 · STS · Dutch, Bilingual Embedding Base scores 77.9 vs 78 — about the same.
Released 2024-06-26. It is the current version in its line.
Yes, the weights are open (Apache-2.0 licence).
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Bilingual Embedding Base is an open-weights embeddings & retrieval model from Lajavaness, released Jun 2024.
EmbeddingGemma 300M (better), STS Multilingual Mpnet Base v2 (open weights).
BetterEmbeddingGemma 300MunstatedOpen+3%—$0.002GOpenSTS Multilingual Mpnet Base v2unstated+18%——Max input: 1K tokens → 0K tokens
BetterQwen 3 Embedding 8BunstatedOpen+12%—$0.01Memory (4-bit): 0.2 GB → 4.5 GBiBetterMultilingual E5 Large InstructunstatedOpen+6%—$0.01Memory (4-bit): 0.2 GB → 0.3 GBiBetterMultilingual E5 LargeunstatedOpen+3%—$0.01Memory (4-bit): 0.2 GB → 0.3 GB
Jina Embeddings v5 Text NanounstatedOpen+11%—$0.02
Llama Embed Nemotron 8BOpen+11%——Memory (4-bit): 0.2 GB → 4.5 GB