Jina Embeddings v2 Small English
Jina Embeddings v2 Small English provides a smaller alternative to the 137-million-parameter base model. It keeps the v2 family's 8,192-token context while reducing the parameter count to 33 million.
How the model works
The JinaBERT backbone was pretrained on C4 and then trained using sentence pairs and hard negatives. Its bidirectional ALiBi position method extends beyond the 512-token sequences used during embedding training. Mean pooling converts token representations into a sentence or document vector.Jina Embeddings v2 Small English model details
Use and access
The weights are available under Apache 2.0. The official card includes local Transformers and Sentence Transformers paths. Small English is a separate trained checkpoint, not a quantized copy of Base English. Its release remains useful for comparing older low-resource retrieval systems, without implying that it offers newer v3 or v5 task adapters.Jina Embeddings v2 Small English model details
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | Downloadable |
| Model identifiers | jina-embeddings-v2-small-enjinaai/jina-embeddings-v2-small-en |
| Context window | 8,192 tokens Maximum input sequence in the model card; serving limits can be lower. Jina Embeddings v2 Small English model details |
| Parameters | 33 millionJina Embeddings v2 Small English model details |
| Weight license | Apache 2.0Jina Embeddings v2 Small English model details |
| Base model | JinaBERT SmallJina Embeddings v2 Small English model details |
Related models
- Jina Embeddings v2 Base English
Jina AI · Jina Embeddings v2
An English long-document encoder with permissively licensed weights.
Launch date Downloadable - Jina Embedding v1 Small English
Jina AI · Jina Embeddings v1
A small English vector model from Jina's first embedding release.
Launch date Downloadable