How it works Rerank 5 Pro?
Rerank 5 Pro is the most accurate model in the Cohere Rerank 5 family: it evaluates document relevance to a query and reorders results before they are fed into a generative model's context. In BotHub, the model works via a unified OpenAI-compatible API alongside over 250 other models. Payment is in rubles on a pay-as-you-go basis without subscriptions or foreign cards, and data is encrypted and not stored. Use cases: second-stage search after vector retrieval in RAG, corporate knowledge base accuracy, selecting snippets for chatbot context, filtering search results, and curating large document sets. Teams have access to contracts, invoices, electronic document management, and admin panel limits.Frequently asked questions about Rerank 5 Pro
You can use generated results for commercial purposes. You own all rights to the content you create. The only restriction: make sure your prompt does not include copyrighted third-party material. You are responsible for respecting the rights to any input data.
rerank-5-pro is a ranking model: you provide a query and a list of texts, and it evaluates the relevance of each, reordering them in descending order. Key tasks include knowledge base search, RAG, filtering output before sending it to a language model, and sorting results in your application.
The model does not write its own text: its response is the order of documents and relevance scores. According to our data, the context is up to 32,768 tokens, and the output limit is 29,491 tokens. If you need a long, coherent response, rank the output using rerank-5-pro and assign the text generation to a language model in BotHub.
Our data does not indicate a separate reasoning mode. The ranker solves a different problem: it matches a query with a document and immediately provides a score, rather than reasoning step-by-step. If you need analysis with explanations or planning, switch to a conversational model in BotHub — the integration will remain the same.
Our data does not indicate function calling or structured JSON output. However, the ranking result is already simple: a list of documents with indices and scores, which is easy to parse with any HTTP client. For complex scenarios involving tools, use a separate model in BotHub.
Our data does not indicate support for images, audio, or video: the model works with 'query-document' text pairs. If you need to recognize images or transcribe recordings, BotHub has other types of models — you can call them from the same dashboard and the same API.
Our data does not contain a separate quality assessment for Russian. It is easy to check: take a dozen of your queries and documents, compare the output order with the expected one — this will show you if the model is suitable for your specific corpus. The test will take minutes, not days.