jina-embeddings-v5-omni-small-retrieval

by jinaai

Retrieval-tuned Jina v5 Omni model for text, image, audio, and video embeddings

13Kdl/month

2Bparams

HuggingFace Use in Pipeline

Identifiers

Model ID

jinaai/jina-embeddings-v5-omni-small-retrieval

Feature URI

mixpeek://image_extractor@v1/jina_embeddings_v5_omni_small_retrieval

Overview

Jina Embeddings v5 Omni Small Retrieval is a retrieval-specific fine-tune of Jina's omni embedding model. It targets unified embedding across text, image, audio, and video while preserving the Sentence Transformers style workflow many retrieval stacks already use.

On Mixpeek, the retrieval variant is a strong candidate for agentic search over mixed media because a single query can fan out across frames, audio segments, captions, and document snippets without changing the retriever shape.

Architecture

Transformer-based omni embedding model with Qwen3-VL-Audio lineage. The retrieval variant is fine-tuned from jinaai/jina-embeddings-v5-omni-small for multimodal search tasks.

Mixpeek SDK Integration

import { Mixpeek } from "mixpeek";

const mx = new Mixpeek({ apiKey: "API_KEY" });

await mx.collections.ingest({
  collection_id: "mixed-media",
  source: { url: "https://example.com/product-demo.mp4" },
  feature_extractors: [{
    feature: "multimodal_embedding",
    model: "jinaai/jina-embeddings-v5-omni-small-retrieval"
  }]
});