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.
#200 of 200 in Embeddings.
On MTEB Mandarin Chinese v1 · Mean over task types · Mandarin Chinese, Ternary Weight Embedding scores 67.6 vs 67.9 — about the same.
Released 2024-10-23. It is the current version in its line.
Yes, the weights are open.
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Ternary Weight Embedding is an open-weights embeddings & retrieval model from malenia1, released Oct 2024.
Qwen 3 Embedding 8B (better), QZhou Embedding Zh (open weights).
BetterQwen 3 Embedding 8BunstatedOpen+11%—$0.01Memory (4-bit): 0.1 GB → 4.5 GBKOpenQZhou Embedding Zh+19%——Memory (4-bit): 0.1 GB → 4.5 GB
BetterYoutu EmbeddingOpen+17%——Memory (4-bit): 0.1 GB → 1.6 GBKBetterQZhou EmbeddingunstatedOpen+16%——Memory (4-bit): 0.1 GB → 4.2 GB