LCO-Embedding-Omni-7B

Name: LCO-Embedding-Omni-7B
Author: LCO-Embedding

by LCO-Embedding

SOTA omni-modal embedding for text, images, audio, and video in one vector space

2.1Kdl/month

7Bparams

HuggingFace Run it on your own data, free

Identifiers

Model ID

LCO-Embedding/LCO-Embedding-Omni-7B

Feature URI

mixpeek://image_extractor@v1/lco_embedding_omni_7b_v1

Overview

LCO-Embedding-Omni-7B is a language-centric omni-modal embedding model that maps text, images, audio, and video into a shared vector space. It achieves state-of-the-art on both the MIEB image embedding benchmark and MAEB audio embedding benchmark: notably reaching audio SOTA without explicit audio training data.

Built on Qwen2.5-Omni-Thinker-7B with a sentence-transformer last-token-pooling head, it demonstrates the 'Generation-Representation Scaling Law': strong generative backbones produce strong embeddings across all modalities.

Architecture

7B parameter model using Qwen2.5-Omni-Thinker as backbone. Employs last-token pooling via sentence-transformers for fixed-dimensional embeddings. Cross-modal alignment enables retrieval across modality boundaries without modality-specific heads.

Mixpeek SDK Integration

// No extractor parameter takes a Hugging Face model id (checked against
// GET /v1/discovery/extractors, which returns 13), so LCO-Embedding-Omni-7B runs
// on your side and the output is upserted through POST
// /v1/namespaces/{namespace_id}/documents/upsert. On Enterprise the other
// path is to upload the weights instead: POST /v1/namespaces/{id}/models
// accepts the huggingface format and a custom plugin loads them.
const res = await fetch(
  "https://api.mixpeek.com/v1/namespaces/ns_your_namespace/documents/upsert",
  {
    method: "POST",
    headers: {
      Authorization: "Bearer API_KEY",
      "Content-Type": "application/json",
    },
    body: JSON.stringify({
      collection_id: "col_your_collection",
      documents: [
        {
          document_id: "asset-00412",
          // The vector name has to match a vector index on the collection.
          vectors: { "text-embedding": yourVector },
          payload: { source_key: "archive/2026/asset-00412" },
        },
      ],
    }),
  },
);

// Managed alternative, if this exact model is not the requirement:
// text_extractor@v1 runs intfloat/multilingual-e5-large-instruct
// (1024-d) over a bucket, with no inference of your own.