← All models

Qwen3-Rerank

Published Publisher Alibabaqwen3-rerank

Overview

Qwen3-Rerank is a text reranking model based on a Qwen LLM foundation, served through Alibaba Cloud Model Studio. It ranks candidate documents by their relevance to a query, supports more than 100 languages, and is intended for text retrieval and retrieval-augmented generation (RAG). Alibaba describes its performance as aligned with the open-source Qwen3-Rerank series; that statement alone does not establish the weight availability of this particular hosted model. (help.aliyun.com)

Capabilities and uses

  • Reorders text-search results after initial retrieval and returns the most relevant candidates first.
  • Supports top_n to limit returned results and an English instruct parameter to guide the ranking strategy, such as question-answer retrieval or semantic similarity. (alibabacloud.com)

Limits

  • Text input only; it is a reranker rather than a general text-generation model. The model card lists function calling, structured output, web search, context caching, batch inference and fine-tuning as unsupported. (help.aliyun.com)
  • The reranking documentation lists up to 500 documents and 120,000 calculated tokens per request. Its Chinese page lists a 4,000-token maximum per query or document, while the international English page lists 32,768 for Singapore. The model card separately lists a 30,000-token context. These official figures differ; verify the applicable endpoint and region before relying on a limit. (help.aliyun.com)

Release

Alibaba's release log lists the mainland China service on October 21, 2025, and the international deployment on January 27, 2026. (help.aliyun.com)

Pricing

Official Model Studio list prices checked November 17, 2026. Input tokens are billed; output is free. Regional prices and accounts are distinct and are not currency conversions. (help.aliyun.com)

  • China site, Beijing: CNY 0.5 per million input tokens. The pricing page lists a free quota of 1 million tokens, valid for 90 days from the latest of activation, model release or approval. (help.aliyun.com)
  • China site, Singapore, international deployment: CNY 0.74942 per million input tokens. The model card separately displays CNY 0.749 per million. (help.aliyun.com)
  • International site, Singapore, international deployment: USD 0.10 per million input tokens, with a listed free quota of 1 million tokens under the same 90-day timing rule. (alibabacloud.com)

Official sources