Jina models

Serverless Preview Stack Planned

This page collects all Jina models you can use as part of the Elastic Stack. Currently, the following models are available as built-in models:

• jina-embeddings-v3
• jina-reranker-v2

The jina-embeddings-v3 is a multilingual dense vector embedding model that you can use through Elastic Inference Service (EIS). It provides long-context embeddings across a wide range of languages without requiring you to configure, download, or deploy any model artifacts yourself. As the model runs on EIS, Elastic's own infrastructure, no ML node scaling and configuration is required to use it.

The jina-embedings-v3 model supports input lengths of up to 8192 tokens and produces 1024-dimension embeddings by default. It uses task-specific adapters to optimize embeddings for different use cases (such as retrieval or classification), and includes support for Matryoshka Representation Learning, which allows you to truncate embeddings to fewer dimensions with minimal loss in quality.

For more information about the model, refer to the model card on Hugging Face.

Dense vector embeddings are fixed-length numerical representations of text. When you send text to an EIS inference endpoint that uses jina-embeddings-v3, the model returns a vector of floating-point numbers (for example, 1024 values). Texts that are semantically similar have embeddings that are close to each other in this vector space. Elasticsearch stores these vectors in dense_vector fields or through the semantic_text type and uses vector similarity search to retrieve the most relevant documents for a given query. Unlike ELSER, which expands text into sparse token-weight vectors, this model produces compact dense vectors that are well suited for multilingual and cross-domain use cases.

To use jina-embeddings-v3, you must have the appropriate subscription level or the trial period activated.

Create an inference endpoint that references the jina-embeddings-v3 model in the model_id field.

				PUT _inference/text_embedding/eis-jina-embeddings-v3
					{
  "service": "elastic",
  "service_settings": {
    "model_id": "jina-embeddings-v3"
  }
}
		

The created inference endpoint uses the model for inference operations on Elastic Inference Service. You can reference the inference_id of the endpoint in text_embedding inference tasks or search queries. For example, the following API request ingests the input text and produce embeddings.

				POST _inference/text_embedding/eis-jina-embeddings-v3
					{
  "input": "The sky above the port was the color of television tuned to a dead channel.",
  "input_type": "ingest"
}
		

• jina-embeddings-v3 works best on small, medium or large sized fields that contain natural language. For connector or web crawler use cases, this aligns best with fields like title, description, summary, or abstract. Although jina-embeddings-v3 has a context window of 8192 tokens, it's best to limit the input to 2048-4096 tokens for optimal performance. For larger fields that exceed this limit - for example, body_content on web crawler documents - consider chunking the content into multiple values, where each chunk can be under 4096 tokens.
• Larger documents take longer at ingestion time, and inference time per document also increases the more fields in a document that need to be processed.
• The more fields your pipeline has to perform inference on, the longer it takes per document to ingest.

jina-reranker-v2 is a multilingual cross-encoder model that helps you to improve search relevance across over 100 languages and various data types. The model significantly improves information retrieval in multilingual environments. jina-reranker-v2 is available out-of-the-box and supports Elastic deployments using the Elasticsearch Inference API. You can use the model to improve existing search applications like hybrid semantic search, retrieval augmented generation (RAG), and more. You can use the model through Elastic Inference Service (EIS), Elastic's own infrastructure, without the need of managing infrastructure and model resources.

For more information about the model, refer to the model card on Hugging Face.

To use jina-reranker-v2, you must have the appropriate subscription level or the trial period activated.

Create an inference endpoint that references the jina-reranker-v2 model in the model_id field.

				PUT _inference/rerank/eis-jina-reranker-v2
					{
  "service": "elastic",
  "service_settings": {
    "model_id": "jina-reranker-v2"
  }
}
		

The created inference endpoint uses the model for inference operations on Elastic Inference Service. You can reference the inference_id of the endpoint in rerank inference tasks. For example, the following API request ingests the input strings and ranks them by relevance:

				POST _inference/rerank/eis-jina-reranker-v2
					{
  "input": ["luke", "like", "leia", "chewy","r2d2", "star", "wars"],
  "query": "star wars main character"
}