bge-m3

Name: bge-m3
Author: BAAI

by BAAI

Hybrid retrieval in one model -- dense, sparse, and ColBERT embeddings from a single forward pass

36.1Mdl/month

3,439likes

568Mparams

HuggingFace Run it on your own data, free

Identifiers

Model ID

BAAI/bge-m3

Feature URI

mixpeek://text_extractor@v1/baai_bge_m3_v1

Overview

BGE-M3 is BAAI's multi-functionality embedding model that produces three types of embeddings simultaneously: dense vectors for semantic search, sparse vectors for lexical matching, and ColBERT-style multi-vector representations for fine-grained late interaction. This eliminates the need to run separate models for different retrieval strategies.

The model supports 100+ languages and handles up to 8192 tokens of input, making it suitable for long documents. On Mixpeek, BGE-M3 powers hybrid retrieval pipelines where a single ingest pass produces all three representation types, and the retriever fuses them at query time for higher recall than any single strategy alone.

Architecture

XLM-RoBERTa backbone, 568M parameters. Produces dense embeddings (1024d), sparse term-weight vectors, and ColBERT multi-vector representations from one forward pass. Trained with self-knowledge distillation across 100+ languages.

Mixpeek SDK Integration

// No extractor parameter takes a Hugging Face model id (checked against
// GET /v1/discovery/extractors, which returns 13), so bge-m3 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.