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.
#131 of 200 in Embeddings.
On MTEB Indic v1 · Bitext Mining · multilingual (20 languages), KaLM Embedding Multilingual Mini Instruct v1 scores 49.6 vs 50.8 — lower.
Released 2024-10-23. 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.
KaLM Embedding Multilingual Mini Instruct v1 is an open-weights embeddings & retrieval model from HIT-TMG, released Oct 2024.
Multilingual E5 Large Instruct (better).
BetterQwen 3 Embedding 0.6BunstatedOpen+36%—$0.01Memory (4-bit): 0.3 GB → 0.4 GBVBetterVoyage 3 Liteunstated+22%—$0.02Max input: 33K tokens → 32K tokensiBetterE5 Mistral 7B InstructunstatedOpen+19%—$0.022Memory (4-bit): 0.3 GB → 4.3 GB
Harrier OSS v1 0.6BunstatedOpen+62%——Memory (4-bit): 0.3 GB → 0.4 GBcF2LLM v2 8BunstatedOpen+61%——Memory (4-bit): 0.3 GB → 4.5 GBOSolon Embeddings Large 0.1unstatedOpen+59%——Max input: 33K tokens → 1K tokenscF2LLM v2 4BunstatedOpen+58%——Memory (4-bit): 0.3 GB → 2.4 GB