Jina Embeddings v5 Text Small
Text Small is the 677-million-parameter member of Jina's fifth embedding generation, released in February 2026. It turns multilingual text into vectors for search, classification, clustering and similarity matching.
How the model works
Jina trained a Qwen3-0.6B-Base backbone with knowledge distilled from Qwen3-Embedding-4B and task-specific contrastive losses. Four adapters share that backbone. Last-token pooling produces 1,024-dimensional vectors; Matryoshka training permits shorter vectors down to 32 dimensions. The full input limit is 32,768 tokens.Jina Embeddings v5 Text Small model details
Use and access
The base download includes all four adapters. Separate retrieval, text-matching, classification and clustering repositories merge one adapter into the same backbone for serving tools. These are task variants, not four independent foundation models. Jina documents its API, Elastic Inference Service and local weights. The downloadable weights use CC BY-NC 4.0; commercial deployment has separate terms.Jina Embeddings v5 Text Small model detailsJina Embeddings v5 Text Small technical overview
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | API, Downloadable |
| Model identifiers | jina-embeddings-v5-text-smalljinaai/jina-embeddings-v5-text-smalljina-embeddings-v5-text-small-retrievaljina-embeddings-v5-text-small-text-matchingjina-embeddings-v5-text-small-classificationjina-embeddings-v5-text-small-clustering |
| Context window | 32,768 tokens Maximum input sequence in the model card; serving limits can be lower. Jina Embeddings v5 Text Small model details |
| Parameters | 677 million including task adaptersJina Embeddings v5 Text Small model details |
| Embedding dimensions | 1,024; Matryoshka sizes down to 32Jina Embeddings v5 Text Small model details |
| Weight license | CC BY-NC 4.0Jina Embeddings v5 Text Small model details |
| Base model | Qwen3-0.6B-BaseJina Embeddings v5 Text Small model details |
Related models
- Jina Embeddings v5 Omni Small
Jina AI · Jina Embeddings v5 Omni
A shared search encoder for text, images, audio, video and document files.
Launch date API · Downloadable - Jina Embeddings v5 Text Nano
Jina AI · Jina Embeddings v5
A EuroBERT-based text encoder for search with a smaller memory footprint.
Launch date API · Downloadable