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.
#12 of 27 in Multimodal.
On MVEB text video v1 · Video Zero-shot Classification · English, Pe Av Base scores 49 vs 50 — about the same.
Released 2025-12-22. 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.
Pe Av Base is an open-weights embeddings & retrieval model from Meta, released Dec 2025.
Jina Embeddings v5 Omni Small (better), Qwen 3 VL Embedding 8B (open weights).
BetterJina Embeddings v5 Omni SmallunstatedOpen+9%—$0.05Memory (4-bit): 0.6 GB → 1 GB
OpenQwen 3 VL Embedding 8B+21%—$0.09Memory (4-bit): 0.6 GB → 4.9 GBeBetterEbind Audio VisionOpen+18%——Max input: 1K tokens → 0K tokenseBetterEbind FullunstatedOpen+18%——Max input: 1K tokens → 0K tokens+1eBetterEbind Points VisionOpen+18%——Max input: 1K tokens → 0K tokens+1