Discover ANY AI to make more online for less.

Cohere’s Rerank 4 quadruples the context window over 3.5 to cut agent errors and boost enterprise search accuracy

Almost a year after releasing Rerank 3.5, Cohere launched the latest version of its search model, now with a larger context window to help agents find the information they need to complete their tasks. Cohere said in a blog post that Rerank 4 has a 32K context window, representing a four-fold increase compared to 3.5. “This enables the model to handle longer documents, evaluate multiple passages simultaneously and capture relationships across sections that shorter windows would miss,” according to the blog post. “This expanded capacity, therefore, improves ranking accuracy for realistic document types and increases confidence in the relevance of retrieved results.”

Rerank 4 comes in two flavors: Fast and Pro. As a smaller model, Fast is best suited for use cases that require both speed and accuracy, such as e-commerce, programming, and customer service. Pro is optimized for tasks that require deeper reasoning, precision, and analysis, such as generating risk models and conducting data analysis. Enterprise search gained greater importance this year, especially as AI agents have to access more information and context about the organization they work for. Cohere said rerankers “significantly enhance the accuracy of enterprise AI search by refining initial retrieval results.” Rerank 4 addresses the nuance gap created by some bi-encoder embeddings — models that help make retrieval augmented generation (RAG) tasks easier — by using a cross-encoder architecture “that processes queries and candidates jointly, capturing subtle semantic relationships and reordering results to surface the most relevant items,” Cohere said.Performance and benchmarks Cohere benchmarked the models against other reranking models, such as Qwen Reranker 8B, Jina Rerank v3 from Elasticsearch, and MongoDB’s Voyage Rerank 2.5, across tasks in the finance, healthcare, and manufacturing domains. Rerank 4 performed strongly, if not outperformed, its competitors. 
Rerank 3.5 stood out because of its ability to support several languages, and Cohere said Rerank 4 continues that trend. It understands over 100 languages, including state-of-the-art retrieval in 10 major business languages.Agents and reranking models Rerank 4 aims to make agentic tasks understand which data is best suited to their tasks and to provide more context. Cohere noted that the model is a key component of its agentic AI platform, North, as it “integrates seamlessly into existing AI search solutions, including hybrid, vector and keyword-based systems, with minimal code changes.”As more enterprises look to use agents for research and insights, as evidenced by the rise of Deep Research features, models that help filter irrelevant content, such as rerankers, become more essential. “This is especially impactful for agentic AI, where complex, multi-step interactions can quickly drive up model calls and saturate context windows,” Cohere said.The company argues that Rerank 4 helps reduce token usage and the number of retries an agent needs to get things right by preventing low-quality information from reaching the LLM. Self-learning
Cohere said Rerank 4 stands out not just for its strong reranking abilities, but also for being the first reranking model that self-learns. Users can customize Rerank 4 for use cases they encounter more frequently without any additional annotated data. Much like foundation models like GPT-5.2, where people can state preferences and the model remembers these, Rerank 4 users can tell the model their preferred content types and document corpora. If used with Rerank 4 Fast, for example, the model becomes more competitive with larger models because it is more precise and taps specific data users want. “Looking further, we also explored how Rerank 4’s self-learning capability performs on entirely new search domains,” Cohere said. “Using healthcare-focused datasets that mimic a clinician’s need to retrieve patient-specific information — not just expertise from a given medical discipline — we found that enabling Self Learning produced consistent, substantial gains. The result: a clear and significant boost in retrieval quality for Rerank 4 Fast, across the board.”