BGE-VL-base

Name: BGE-VL-base
Author: BAAI

by BAAI

Lightweight vision-language embeddings for image and document retrieval

2Kdl/month

34likes

150Mparams

HuggingFace Run it on your own data, free

Identifiers

Model ID

BAAI/BGE-VL-base

Feature URI

mixpeek://image_extractor@v1/baai_bge_vl_base_v1

Overview

BGE-VL Base is BAAI's compact vision-language embedding model for image-text retrieval and visual document search. It gives teams a smaller open model option when CLIP-style embeddings are too generic and larger multimodal retrievers are unnecessary.

On Mixpeek, BGE-VL Base can index screenshots, product images, scanned pages, and video keyframes so an agent can retrieve visual evidence with natural-language queries before asking a VLM to reason over the result.

Architecture

Sentence Transformers compatible vision-language embedding model with a compact parameter footprint. It maps visual and text inputs into a shared retrieval space for semantic similarity search.

Mixpeek SDK Integration

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

// Managed alternative, if this exact model is not the requirement:
// image_extractor@v1 runs google/siglip-base-patch16-224
// (768-d) over a bucket, with no inference of your own.