677M-parameter multilingual text embedding model scoring 71.7 on MTEB English v2 and 67.7 on MMTEB. Highest performance among multilingual embedding models under 1B parameters. Built on Qwen3-0.6B-Base with distillation from Qwen3-Embedding-4B. Supports 119+ languages, 32K tokens, 1024-dim with Matryoshka truncation. Robust under truncation and binary quantization.
Modalities
Dimensions
1K
Max tokens
33K
Parameters
677M
Price / 1M tokens
—
Type
| MTEB ENGLISH V2 | 71.70 |
| MMTEB | 67.70 |
| Release date | 2026-02-18 |
| License | CC BY-NC 4.0 |
| Model ID | jina-embeddings-v5-text-small |
| Provider | Jina AI |
Get detailed specifications for Jina Embeddings v5 Text Small, including output dimensionality of 1024 dimensions, maximum input token length, supported input modalities, and pricing per million tokens. This embedding model is designed for semantic search, text classification, clustering, and retrieval-augmented generation applications where understanding the relationship between texts is essential.
Review the pricing structure for Jina Embeddings v5 Text Small and compare it against other embedding models from Jina AI and competitors. Understanding embedding model costs is essential when scaling vector search and RAG applications to millions of documents. We provide transparent pricing to help you budget effectively.
Explore the ideal use cases for Jina Embeddings v5 Text Small. Whether you are building a semantic search engine, recommendation system, document classification pipeline, or multilingual retrieval system, understanding this model's capabilities and dimensionality will help you choose the right embedding strategy for your project.