Rerank 5

Cohere's reranker family in Pro and Fast tiers, with optional passage-level Context Focus and calibrated relevance scores.

Cohere · Data, training, and fine-tuning platforms (datasets, labeling, RAG/vector DBs, fine-tuning-as-a-service) · model

Published in Feed

Overview

Rerank 5 is Cohere's reranking model family, published October 22, 2026 in two tiers: rerank-v5.0-pro for ranking quality and rerank-v5.0-fast for low latency. Both have a 32K-token context window, support more than 100 languages and handle semi-structured text such as JSON, tables and YAML. An optional Context Focus feature, off by default, returns scored character spans (start and end offsets) inside each document in the same Rerank call, so an application can pass only the relevant passages to an LLM. Cohere reports that removing half the retrieved text kept 96% of answer quality across 30 QA datasets; these figures are self-reported.

This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.

Key features

  • Two tiers: rerank-v5.0-pro and rerank-v5.0-fast, both with 32K-token context
  • Context Focus: returns scored spans via the relevant_spans field with threshold and max_spans settings; off by default
  • Cohere-reported: half the retrieved text removed kept 96% of answer quality (30 QA datasets); tokens down 41% and accuracy up from 39% to 52% on 6 deep-search benchmarks
  • Self-reported RCP-nDCG@10: Pro 92.2 on ViDoRe V3 and 90.4 on BEIR; Fast 90.6 and 88.1
  • Available on the Cohere API, Model Vault, Microsoft Foundry, Amazon SageMaker, and private VPC or on-prem; proprietary, no open weights

Use cases

  • Re-ordering retrieval results before they reach an LLM in a RAG pipeline
  • Cutting token spend in search agents by sending only relevant passages
  • Multilingual and semi-structured document search (JSON, tables, YAML)

Pricing

API: $0.05 per 1M tokens (Pro) and $0.02 per 1M tokens (Fast). Model Vault: $4.00 per instance-hour for both tiers on the Medium tier.

Pricing reflects the source record and may change. Confirm current terms with the provider.

Dates and sources

Released
22 Oct 2026
Record updated
1 Nov 2026
Record last checked
1 Nov 2026

Explore the Feed

Other entries filed under data, training, and fine-tuning platforms (datasets, labeling, rag/vector dbs, fine-tuning-as-a-service); inclusion does not imply equivalent features.

  • DigitalOcean Agent Data Services

    DigitalOcean's managed data layer for agents: ingestion flows, a catalog with vectors, and SQL or vector query, in private preview.

  • pplx-embed-v2-late

    Perplexity's late-interaction multimodal embedding models (0.6B and 9B) with 128-dim per-token vectors, open on Hugging Face.

  • EmbeddingGemma 2

    Open-weight 740M-parameter multimodal embedding model built on Gemma 4 that encodes text, images, audio, and video for on-device search.

Browse the Feed