BGE Reranker v2 M3

Name: BGE Reranker v2 M3
Author: BAAI

A multilingual cross-encoder reranker from BAAI that reorders search results by relevance, a common second stage in RAG pipelines.

Run it locally

Memory: Runs on CPU; a small GPU speeds it up
Formats: Safetensors (Hugging Face)

Overview

Rerankers read the query and each candidate document together and output a relevance score. This model is built on BGE-M3 and supports many languages.

Highlights

568M parameters
Multilingual
Cross-encoder relevance scores
Apache 2.0 license

Typical uses

Improving retrieval quality in RAG and search.

Things to consider

Slower than embeddings because each query-document pair is scored.

Specifications are taken from the model card on Hugging Face and the provider’s announcement. Benchmark results depend on evaluation settings; see how AI benchmarks work.

#Embedding Models #Open Weights #BAAI

Similar models

View all

Sentence TransformersAug 30, 2021

all-MiniLM-L6-v2

A tiny, fast sentence-embedding model from the Sentence Transformers project, still one of the most downloaded models for semantic search.

Type: Embedding
Context: 512
Weights: Open

BAAISep 12, 2023

BGE Large EN v1.5

BAAI's English embedding model that ranked at the top of MTEB when released and remains a dependable default for retrieval.

Type: Embedding
Context: 512
Weights: Open

BAAIJan 30, 2024

BGE-M3

Open multilingual embedding model from the Beijing Academy of AI supporting dense, sparse and multi-vector retrieval in over 100 languages.

Type: Embedding
Context: 8K
Weights: Open