BGE Reranker Base
Xenova/bge-reranker-base
published Sep 2023 · updated Jun 2025
BGE Reranker Base is a cross-encoder reranking model that scores the relevance of query-document pairs for improved search and retrieval.
specs
| Task | Reranking |
| Architecture | XLM-RoBERTa |
| Parameters | 278M |
| License | MIT |
| Languages | English, Chinese |
about this model
Xenova/bge-reranker-base is a cross-encoder reranker model that scores query-document pairs for relevance, enabling reordering of candidate lists from an initial retrieval step.
Overview
Built on the XLM-RoBERTa architecture (278 million parameters), this model is part of the BGE (BAAI General Embedding) family and supports both English and Chinese. It is optimized for reranking tasks, offering improved performance over earlier BGE rerankers, with support for longer input lengths and broader language coverage.
Benchmark Performance
The following scores are from the MTEB reranking benchmark (MAP and MRR):
| Dataset | MAP | MRR |
|---|---|---|
| CMedQAv1 | 81.27 | 84.14 |
| CMedQAv2 | 84.10 | 86.79 |
| MMarcoReranking | 35.46 | 34.60 |
| T2Reranking | 67.28 | 77.13 |
Additional Details
- License: MIT
- Model size: 278M parameters
- Library: sentence-transformers
- Creation date: 2023-09-11 (last updated 2024-06-24)
This model is hosted by gigarouter as an OpenAI-compatible API, requiring no local installation. Simply call the endpoint with query-document pairs to obtain relevance scores.
best for
- ·Re-ranking initial search results to improve top-k precision
- ·Enhancing retrieval-augmented generation (RAG) pipelines
- ·Answering questions by scoring candidate passages
FAQ
It expects a query and a document pair, returning a relevance score (usually between 0 and 1).
BGE Reranker Base has 278M parameters, which is smaller and faster than the large variant (1.1B parameters).
It is released under the MIT license, allowing free use, modification, and distribution.
Use the OpenAI-compatible endpoint with your API key and specify the model as "Xenova/bge-reranker-base".
Yes, it supports both English and Chinese.
We're benchmarking and onboarding BGE Reranker Base as a hosted, OpenAI-compatible API. Sign in for free credit and be ready when it lands, or tell us you want it and we'll prioritize it.