Qwen3 Reranker 4B is a text reranking model built on the dense foundation of Alibaba's Qwen3 family and released as part of the Qwen3 Embedding and reranking series alongside 0.6B and 8B siblings. With roughly four billion parameters, it inherits the multilingual, long-text, and reasoning capabilities of the underlying Qwen3 base model and is designed to score and reorder candidate passages for retrieval pipelines. The model supports more than 100 languages, including programming languages, making it suitable for cross-lingual and code-aware search scenarios where a first-stage retriever returns many candidates that need fine-grained relevance judgment.

In practical deployments, Qwen3 Reranker 4B serves as a second-stage reranker that can be paired with the Qwen3 Embedding models to combine fast vector retrieval with deeper relevance scoring, and the series targets text retrieval, code retrieval, classification, clustering, and bitext mining tasks. The broader series has demonstrated strong retrieval quality, with the largest 8B embedding variant reported as ranking first on the MTEB multilingual leaderboard at a score of 70.58 as of early June 2025, suggesting that the reranking half of the family is positioned for production-grade multilingual search and ranking workflows. Developers can also supply task- or language-specific instructions to tailor reranking behavior to particular domains.