Qwen3-Reranker-8B

Qwen3-Reranker-8B is a reranking model: it does not search, it re-scores. A search engine returns a rough list of candidate documents, and the reranker reads each query-document pair in full and puts the list in a better order. Alibaba released it on 5 June 2025 alongside the Qwen 3 embedding models, in three sizes meant to be mixed and matched with the embedding side. It leads the family on Chinese retrieval (77.45), multilingual retrieval (72.94) and long documents (70.19), and it is the pairing the maker intends for its top-ranked 8B embedding model. It does not lead everywhere: the 4B variant scores higher on English retrieval and roughly twice as high on following written ranking instructions, which is the kind of detail that decides deployments and which the headline of the release does not mention. The model is built on the Qwen3-8B-Base checkpoint, handles a 32k-token context and over 100 languages including programming languages, and accepts a written instruction describing what "relevant" means for the task at hand — the maker measures a 1 to 5 percent gain from using one, and advises writing it in English even for other languages, because the training instructions were English. Weights are Apache 2.0, commercial use included; the repository records about 0.22M downloads in the past thirty days.

▸Videos

No videos yet.

▸Reviews