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.
#127 of 200 in Embeddings.
On MTEB Polish v1 · STS · Polish, Ivysaur scores 58.3 vs 58.4 — about the same.
Released 2024-04-27. It is the current version in its line.
Yes, the weights are open (MIT licence).
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Ivysaur is an open-weights embeddings & retrieval model from Mihaiii, released Apr 2024.
Multilingual E5 Large Instruct (better), Jina Embeddings v3 (open weights).
OpenJina Embeddings v3unstated+25%—$0.05Memory (4-bit): 0 GB → 0.3 GBiBetterE5 Mistral 7B InstructunstatedOpen+19%—$0.022Memory (4-bit): 0 GB → 4.3 GB
BetterBge Multilingual Gemma2unstatedOpen+21%—$0.11Memory (4-bit): 0 GB → 5.5 GBLBetterBilingual Embedding LargeunstatedOpen+22%——Memory (4-bit): 0 GB → 0.3 GB
Gte Qwen2 7B InstructunstatedOpen+21%——Memory (4-bit): 0 GB → 4.2 GBsMmlw Roberta BaseunstatedOpen+21%——Memory (4-bit): 0 GB → 0.1 GB