netease-youdao/BCEmbedding

What is it?

Key features

• 01Embedding model for Chinese, English, and cross-lingual semantic retrieval.
• 02Reranker model supporting Chinese, English, Japanese, and Korean.
• 03Reranker handles passages longer than 512 tokens, up to 32k tokens.
• 04Reranker outputs relevance scores that can be used to filter low-quality passages.
• 05Embedding model works without task-specific query instructions.
• 06Integrations for LangChain and LlamaIndex, plus direct use via transformers and sentence-transformers.

Use cases

• Building a bilingual Chinese-English knowledge base or document Q&A system.
• Adding a reranking stage to an existing RAG pipeline to improve retrieved context.
• Running cross-lingual search where queries and documents are in different languages.
• Benchmarking embedding and reranker models on RAG tasks using the provided evaluation scripts.

Quick start

1. 1

   Create and activate a conda environment

   conda create --name bce python=3.10 -y
2. 2

   Activate the conda environment

   conda activate bce
3. 3

   Install BCEmbedding from PyPI (install a matching torch first)

   pip install BCEmbedding==0.1.5

Things to know

• !Installing torch manually for your CUDA version is needed to avoid conflicts.
• !The README only lists Chinese and English for embeddings, with the reranker covering Japanese and Korean too.

In collections

Topics

FAQ

What is BCEmbedding?

Bilingual Chinese-English embedding and reranker models from NetEase Youdao, built for retrieval in RAG applications. BCEmbedding provides two models for retrieval pipelines: an EmbeddingModel that turns text into semantic vectors for search and question answering, and a RerankerModel that rescores candidate passages so the most relevant results rise to the top. The models target Chinese, English, and cross-lingual retrieval, and the reranker also covers Japanese and Korean.

Is BCEmbedding free to use? What license does it have?

BCEmbedding is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.

Is BCEmbedding still maintained?

Updated within 6 months. The last commit to BCEmbedding was on Sep 3, 2026.

What can I use BCEmbedding for?

Building a bilingual Chinese-English knowledge base or document Q&A system. Adding a reranking stage to an existing RAG pipeline to improve retrieved context. Running cross-lingual search where queries and documents are in different languages. Benchmarking embedding and reranker models on RAG tasks using the provided evaluation scripts.

How do I install BCEmbedding?

Create and activate a conda environment: conda create --name bce python=3.10 -y Activate the conda environment: conda activate bce Install BCEmbedding from PyPI (install a matching torch first): pip install BCEmbedding==0.1.5

Where can I get BCEmbedding?

The source code and installation instructions are on GitHub: https://github.com/netease-youdao/BCEmbedding