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.
#57 of 89 in Multimodal.
On MIEB English · Image Clustering · English, Siglip Large Patch16 384 scores 79.9 vs 80.2 — about the same.
Released 2024-01-08. 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.
Siglip Large Patch16 384 is an open-weights embeddings & retrieval model from Google, released Jan 2024.
Jina Embeddings v5 Omni Small (better), EVA02 CLIP B 16 (open weights).
BetterJina Embeddings v5 Omni SmallunstatedOpen+8%—$0.05Memory (4-bit): 0.4 GB → 1 GBQOpenEVA02 CLIP B 16unstated+16%——Memory (4-bit): 0.4 GB → 3 GBVBetterVoyage Multimodal 3+3%—$0.12lBetterCLIP ViT L 14 DataComp.XL S13B B90KOpen+8%——lBetterCLIP ViT BigG 14 Laion2B 39B B160kOpen+7%——Memory (4-bit): 0.4 GB → 1.5 GB