← All models

GTE-Rerank-V2

Published Publisher Alibabagte-rerank-v2

Model overview

GTE-Rerank-V2 is a multilingual text reranking model developed by Tongyi Lab. It reorders documents based on their semantic relevance to a query and is suitable for scenarios such as semantic search and retrieval-augmented generation (RAG). (help.aliyun.com)

Capabilities and uses

  • Supports more than 50 languages, including Chinese, English, Japanese, Korean, and Thai.
  • Can rescore candidate documents retrieved in an initial search and return a relevance ranking; top_n can be used to limit the number of results returned. (help.aliyun.com)

Limitations

  • Input is text; each request can contain up to 500 candidate documents, each query or document can contain up to 4,000 tokens, and each request can contain up to 30,000 tokens. The request token count is calculated as “query token count × number of documents + total token count of all documents”; an error is returned when an individual input exceeds the limit, with no automatic truncation. (help.aliyun.com)
  • Does not support function calling, web search, context caching, batch inference, or model tuning. (help.aliyun.com)
  • The official reranking documentation notes that the gte-rerank model will be discontinued on May 30, 2026, and recommends using qwen3-rerank; this notice does not explicitly state the full ID gte-rerank-v2, and the documentation still lists v2. Actual availability should be checked in the console. (alibabacloud.com)

Release date

March 20, 2025 is the date the model was made available in the Alibaba Cloud Model Studio China (Beijing) region, and does not necessarily represent the date of its initial development or release. (docs.bailian.console.aliyun.com)

Pricing

The following are standard prices listed on official pages. All charges are based on input tokens; output is not charged. Quotes for accounts on different platforms should not be treated as the same plan or directly converted.

Official sources