Cohere Releases Rerank 5 with Passage-Level Context Focus and Calibrated Scores

Cohere has launched Rerank 5, the latest version of its retrieval reranking model, featuring improved passage-level context focus and better-calibrated relevance scores. The calibrated scoring is a meaningful improvement for RAG pipelines where downstream logic — such as thresholds for answer generation or citation inclusion — depends on scores being interpretable and consistent across queries. Passage-level context focus means the model is better at distinguishing relevance within long documents rather than treating chunks as atomic units, which directly improves recall precision in complex retrieval tasks. For developers building production RAG systems, Rerank 5 is a drop-in upgrade that can meaningfully reduce hallucination rates caused by poorly ranked retrieved context. This release continues Cohere's positioning as the retrieval-focused enterprise AI provider.