Cross-encoder reranking
According to the model card, the reranker takes question and document as input and directly outputs similarity instead of an embedding.
Open Source Model Profile · BAAI
bge-reranker-large
bge-reranker-large is a cross-encoder reranking model from BAAI. According to the model card, it reranks top-k documents returned by embedding models and outputs similarity directly.
- Publisher
- BAAI
- Task
- feature-extraction
- Model type
- xlm-roberta
- License
- mit
- Library
- transformers
- Publication status
- Accepted · not indexed
Model overview
bge-reranker-large is published by BAAI for feature-extraction and text-classification-backed reranking. The captured configuration identifies XLMRobertaForSequenceClassification with model type xlm-roberta. Safetensors metadata reports 559,891,971 parameters, and the model card recommends it for reranking top-k embedding results.
Recorded capabilities
Top-k workflow
According to the model card, a BAAI embedding model retrieves the top 100 documents and the reranker reranks them to obtain the final top results.
Reported comparison table
According to the model card, its table reports an average of 66.09 across the listed reranking sets, above the listed bge-reranker-base average of 65.42.
Transformers and ONNX paths
According to the model card, it can be used with Transformers sequence-classification tooling and with ONNX through optimum.
Use cases in the source record
- Reranking top-k documents retrieved by a BAAI embedding model to produce final top results, as described in the model card.
- Retrieval pipelines that use question-and-document input to score similarity directly instead of computing embeddings.
- Fine-tuning the cross-encoder for domain reranking when embedding-only accuracy is insufficient, as recommended in the card.
Limitations and unknowns
- Reported table scores come from the publisher model card and were not independently verified by Ethen.
- No Ethen-measured evaluation results were extracted from this record.
- No context-window value was extracted.
- Provider state is historical snapshot data, not independently refreshed current availability.
Source and provenance
Source: BAAI/bge-reranker-large
Captured: Unknown. Processed: 2026-09-07T19:34:30.343634+00:00.
We have updated the new reranker , supporting larger lengths, more languages, and achieving better performance. FlagEmbedding Model List | FAQ | Usage | Evaluation | Train | Citation | License More details please refer to our Github: FlagEmbedding . English | 中文 FlagEmbedding focuses on retrieval-augmented LLMs, consisting of the following projects currently: Long-Context LLM : Activation Beacon Fine-tuning of LM : LM-Cocktail Embedding Model : Visualized-BGE , BGE-M3 , LLM Embedder , BGE Embedding Reranker Model : llm rerankers , BGE Reranker Benchmark : C-MTEB News 3/18/2024: Release new rerankers , built upon powerful M3 and LLM…
F001F002F003F004F005F006F007F008F009F010F016F025F026F028F029F032F033F035F037F040F041