How AI describes BAAI/bge-reranker-v2-m3

BAAI's BGE Reranker v2 M3 on Hugging Face is a multilingual text-embedding and reranking model designed to improve dense retrieval and text classification across languages. It uses self-knowledge distillation to produce multi-granularity embeddings and is built to work with the sentence-transformers ecosystem, with deployment via Hugging Face Inference Endpoints (and Azure options). The model is open-source under Apache-2.0, available in safetensors format, and hosted with US-region endpoints for live inference.