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.
#86 of 200 in Embeddings.
On MTEB multilingual v2 · Reranking · multilingual (22 languages), Stella En 1.5B v5 scores 61.4 vs 61.4 — about the same.
Released 2024-07-12. 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.
Stella En 1.5B v5 is an open-weights embeddings & retrieval model from NovaSearch, released Jul 2024.
Qwen 3 Embedding 8B (better), F2LLM v2 8B (open weights).
BetterQwen 3 Embedding 8BunstatedOpen+7%—$0.01Memory (4-bit): 0.9 GB → 4.5 GBcOpenF2LLM v2 8Bunstated+15%——Memory (4-bit): 0.9 GB → 4.5 GB
BetterJina Embeddings v5 Text NanounstatedOpen+5%—$0.02iBetterE5 Mistral 7B InstructunstatedOpen+4%—$0.022Memory (4-bit): 0.9 GB → 4.3 GB
Llama Embed Nemotron 8BOpen+10%——Memory (4-bit): 0.9 GB → 4.5 GBOOcten Embedding 8BOpen+10%——Memory (4-bit): 0.9 GB → 4.5 GB