Skip to content

jina-clip-v2

Multilingual image and text embeddings

Visit official site

What jina-clip-v2 is

The model is described as a Jina AI multilingual multimodal embedding model that aligns image and text representations through joint training.

What you can do with it

  • Retrieve images across 89 languages
  • Encode text and 512x512 images
  • Pick an output dimension from 64 to 1024

Getting started

  • Install transformers, einops, timm and pillow
  • Load the model with AutoModel.from_pretrained
  • Call encode_text and encode_image on the data

Pros & cons

✓ What we like

  • Text and visual encoders are jointly trained
  • Commercial use via API and cloud platforms is listed

! What to watch out for

  • Output dimension is chosen from the 64 to 1024 range
  • Use follows the documented encode_text and encode_image calls

FAQ

How many languages are supported?

The material states 89 languages.

Which encoders are named?

Jina-XLM-RoBERTa and EVA02-L14 are listed.

Last reviewed: 2026-11-15

More AI search engine tools

View all →

How we review