MODELS

BAAI/bge-reranker-v2-m3

by BAAI

modellead-sourcehugging-face-popular-modelssource:github.comrerankercross-encodermultilingualbgetext-classificationhugging-faceFlagEmbedding

Visit Website Documentation GitHub

Overview

Multilingual lightweight cross-encoder reranker from BAAI’s BGE reranker v2 family.

Details

BAAI/bge-reranker-v2-m3 is listed on Hugging Face as a text-classification model and is documented in the BGE-Reranker-v2 documentation. The BGE documentation describes it as a multilingual reranker with 568M parameters and a 2.27 GB model size, and shows usage with FlagReranker. The FlagEmbedding repository describes BAAI/bge-reranker-v2-m3 as a multilingual lightweight cross-encoder reranker with strong multilingual capabilities and fast deployment.

When to Use

Use when you need to rerank candidate passages or documents after an initial retrieval step in multilingual search or RAG workflows. Use when you want a BGE reranker v2 model with documented FlagReranker usage and a lighter multilingual cross-encoder profile.

Getting Started

Open the model page on Hugging Face: https://huggingface.co/BAAI/bge-reranker-v2-m3
Review the BGE-Reranker-v2 documentation for model size
parameter count
and FlagReranker example code.
Use the FlagEmbedding GitHub repository as the implementation reference for running BGE reranker models.
Test the reranker on a small set of query-document pairs before using it in a production retrieval pipeline.

Key Features

•Multilingual reranker listed in the BGE-Reranker-v2 documentation
•568M-parameter model with documented 2.27 GB size
•Lightweight cross-encoder reranker according to the FlagEmbedding repository
•Example usage shown with FlagReranker
•Hugging Face model task listed as text-classification

Capabilities

•multilingual reranking
•cross-encoder reranking
•text classification
•retrieval result reranking

Last updated Jun 4, 2026