jina-embeddings-omni-small-retrieval

About

What jina-embeddings-omni-small-retrieval is

jina-embeddings-v5-omni-small-retrieval: Retrieval-Targeted Omni Embedding (Small) ArXiv | Blog Average score vs. parameter count across image (MIEB-Lite), video (MMEB-V), and audio (MAEB) benchmarks — jina-v5-omni-nano and jina-v5-omni-small define the open-weight frontier (Table 1 in the ArXiv report). Model Overview jina-embeddings-v5-omni-small-retrieval is a multimodal embedding model that accepts text, images, video, and audio and produces embeddings in a shared vector space aligned with the text-only jinaai/jina-embeddings-v5-text-small-retrieval — so you can index with text and query with any modality, no reindexing. For a more compact alternative, see jinaai/jina-embeddings-v5-omni-nano-retrieval. This is the retrieval-targeted variant of the jina-embeddings-v5-omni-small family. The combined repo containing all task adapters is jinaai/jina-embeddings-v5-omni-small. Via Elastic Inference Service The fastest way to use v5-omni in production. Elastic Inference Service (EIS) provides managed embedding inference with built-in scaling, so you can generate embeddings directly within your Elastic deployment. See the Elastic Inference Service documentation for setup details. Install For minimum versions see the Requirements section below (transformers = 4.57, torch = 2.5; vLLM path validated with vllm == 0.20.1). Quickstart Retrieval needs to know which side of the pair you are encoding. The raw model.embed(...) path above uses explicit Query: / Document: text prefixes, and the model.encode(..., promptname=...) path handles the prefix for you. Via sentence-transformers (below) the same intent is expressed with encodequery() / encodedocument() helpers. A bare encode(text) without a prefix or promptname is ambiguous and will not match either retrieval side cleanly. This applies to every modality, not just text: to encode an image, video, or audio clip as a query or document, either prepend the same Query: / Document: prefix to the text alongside the media placeholder on the raw path (e.g. text="Query: <|visionstart|<|imagepad|<|visionend|"), or pass the media straight to encodequery(...) / encodedocument(...) via sentence-transformers. No dtype, device, minpixels

Summarised from the published model card. Read the full card on the HuggingFace links below.

Specifications

What it is

How it works

How embedding models work

Variants

Sizes & precisions

Open weights ship in several sizes and precisions. One page, all the variants — pick the one that fits your GPU. VRAM figures are estimates from model size.

Using it via the API

Call it like any OpenAI endpoint

Once AxForge deploys jina-embeddings-omni-small-retrieval for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (jina-embeddings-omni-small-retrieval below is illustrative; you get the exact model name on deployment.)

$ curl -sS https://api.axforge.ai/v1/embeddings \
  -H "Authorization: Bearer $AXFORGE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"jina-embeddings-omni-small-retrieval","input":"text to embed"}'

Details

Languages, data & research

Papers

Licence

Commercial licence needed

Sources

Weights & code

Explore

More embedding models