GME-VARCO-VISION-Embedding

Name: GME-VARCO-VISION-Embedding
Author: NCSOFT

by NCSOFT

Multimodal image and video embedding model for vision-heavy retrieval

214dl/month

12likes

8.3Bparams

HuggingFace Run it on your own data, free

Identifiers

Model ID

NCSOFT/GME-VARCO-VISION-Embedding

Feature URI

mixpeek://image_extractor@v1/ncsoft_gme_varco_vision_embedding_v1

Overview

GME VARCO Vision Embedding is NCSOFT's multimodal embedding model for image, text, and video retrieval. It is based on Qwen2-VL-7B-Instruct and is tagged for video embedding and feature extraction.

On Mixpeek, it fits archives where agents need to search visual scenes and short clips by text, then pass the retrieved moments into a captioner, VLM, or workflow tool.

Architecture

Fine-tuned Qwen2-VL-7B-Instruct model for multimodal embedding. The model card lists image-text-to-text and feature-extraction tags, plus video embedding support.

Mixpeek SDK Integration

// No extractor parameter takes a Hugging Face model id (checked against
// GET /v1/discovery/extractors, which returns 13), so GME-VARCO-VISION-Embedding 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: { "multimodal-embedding": yourVector },
          payload: { source_key: "archive/2026/asset-00412" },
        },
      ],
    }),
  },
);

// Managed alternative, if this exact model is not the requirement:
// multimodal_extractor@v2 runs google/gemini-embedding-2
// (3072-d) over a bucket, with no inference of your own.