Qwen / qwen/qwen3-reranker-8b

Qwen3 Reranker 8B - access through LLMTR

Qwen3 Reranker 8B evaluates a query together with a list of candidate documents and produces a relevance score between 0 and 1 for each one. Embedding search vectorizes query and document separately for speed; a reranker reads both at once and therefore orders results more accurately. The typical pattern is to retrieve the top 50-100 candidates with an embedding model, rerank them, and pass only the most relevant few to the language model, which improves answer quality and lowers the generation model's token cost. It reads over 100 languages, including Turkish, and accepts up to 32,768 tokens per query-document pair. It is called through /v1/rerank and produces no text. Billing is on total processed tokens: query and document tokens are counted together for every pair.