Cohere announced the launch of Rerank 5 in Pro and Fast tiers, introducing a Context Focus feature to trim document passages and reduce token costs.

{
  "signal_id": "dad5eb41-7c7b-5cc0-adfc-c0f1df3f3b40",
  "signal_type": "reddit-company",
  "signal_subtype": "productLaunch",
  "detected_at": "2026-10-21T15:09:54+00:00",
  "company": {
    "name": "Cohere",
    "domain": "cohere.com"
  },
  "data": {
    "nsfw": false,
    "stage": "confirmed",
    "awards": 0,
    "timing": "completed",
    "topics": [
      "ai",
      "reranking",
      "llm",
      "product launch"
    ],
    "post_id": "1xbprpo",
    "summary": "Cohere announced the launch of Rerank 5 in Pro and Fast tiers, introducing a Context Focus feature to trim document passages and reduce token costs.",
    "comments": [
      {
        "url": "https://www.reddit.com/r/LangChain/comments/1xbprpo/comment/19rfwrqwj54bm/",
        "depth": 0,
        "score": 2,
        "author": "omniconflux",
        "excerpt": "Does Context Focus allow dynamic adjustment per query, or is it a fixed setting? Curious how it handles multi-hop reasoning across long enterprise docs compared to standard chunking.",
        "posted_at": "2026-10-21T16:09:55.000Z",
        "author_url": "https://www.reddit.com/user/omniconflux/"
      },
      {
        "url": "https://www.reddit.com/r/LangChain/comments/1xbprpo/comment/19rjgagxhtov8/",
        "depth": 0,
        "score": 2,
        "author": "Brain-NotFound2126",
        "excerpt": "The context-selection part is interesting, especially for long documents where sending everything gets expensive. The tricky bit seems to be making sure the passages that get dropped aren't needed to interpret the ones that remain. How is this trade-off tested on real documents rather than clean benchmark examples?",
        "posted_at": "2026-10-23T15:48:56.000Z",
        "author_url": "https://www.reddit.com/user/Brain-NotFound2126/"
      },
      {
        "url": "https://www.reddit.com/r/LangChain/comments/1xbprpo/comment/19rhtdapkxpnm/",
        "depth": 0,
        "score": 2,
        "author": "Most_Populer_AiTools",
        "excerpt": "Context Focus sounds useful for reducing token costs without losing relevant information.",
        "posted_at": "2026-10-22T17:47:55.000Z",
        "author_url": "https://www.reddit.com/user/Most_Populer_AiTools/"
      },
      {
        "url": "https://www.reddit.com/r/LangChain/comments/1xbprpo/comment/19rjctmx89deq/",
        "depth": 0,
        "score": 2,
        "author": "Left-Confidence9155",
        "excerpt": "I wonder how much the results change when you adjust the Context Focus setting. Would be interesting to compare a few settings on the same documents and see where it starts dropping useful details.",
        "posted_at": "2026-10-23T14:31:12.000Z",
        "author_url": "https://www.reddit.com/user/Left-Confidence9155/"
      },
      {
        "url": "https://www.reddit.com/r/LangChain/comments/1xbprpo/comment/19rhse989yl2d/",
        "depth": 0,
        "score": 1,
        "author": "ram-popular",
        "excerpt": "Context Focus sounds useful for cutting token costs. How does it handle answers that span multiple passages, like a table whose meaning depends on a header paragraph several sections earlier? Is there a way to measure the recall tradeoff as you turn the dial down, so we can pick a setting with evidence rather than guesswork?",
        "posted_at": "2026-10-22T17:26:05.000Z",
        "author_url": "https://www.reddit.com/user/ram-popular/"
      },
      {
        "url": "https://www.reddit.com/r/LangChain/comments/1xbprpo/comment/19rj3k0y9wcgw/",
        "depth": 0,
        "score": 1,
        "author": "Impossible_Error_280",
        "excerpt": "This actually sounds pretty useful for long documents. I'm curious though, what happens when a small piece of context seems irrelevant on its own but turns out to be important for the final answer?",
        "posted_at": "2026-10-23T11:03:26.000Z",
        "author_url": "https://www.reddit.com/user/Impossible_Error_280/"
      }
    ],
    "evidence": [
      "[post] Today we’re introducing Rerank 5, and since it’s available in LangChain, I thought i’d put it here.",
      "[post] We’re also introducing a totally new feature to Rerank that (to our knowledge) we’ve never seen in any other reranking model, Context Focus."
    ],
    "virality": "somewhat_high",
    "post_date": "2026-10-21T15:09:54.000Z",
    "post_kind": "multi_media",
    "post_text": "Hi there, El from Cohere here. Today we’re introducing Rerank 5, and since it’s available in LangChain, I thought i’d put it here.\n\nRerank 5 available in Pro and Fast tiers just like Rerank 4 but far outdoes it – along with every other model we tested. Pro is number 1 on the nanoMTEB and ViDoRe V3 benchmarks and Fast outdoes other speedy models like Qwen3-Reranker-0.6B and Qwen3-Reranker-8B (which is a lot bigger).\n\nBoth are also leaders across multilinguality, financial document performance, healthcare docs, and more (I’ll let you read the full blog for all that info).\n\nWe’re also introducing a totally new feature to Rerank that (to our knowledge) we’ve never seen in any other reranking model, Context Focus. Enterprise docs are only getting longer, so more and more context is getting passed to LLMs=more tokens. Context Focus grabs only the specific passages in a document relevant to your query, and it reduces token costs pretty significantly. It’s an opt-in feature and works like a dial, so you can choose exactly how much context you want.\n\nLike I said, it’s available through LangChain (and a bunch of other places like our API). See our blog for more info and lmk what you think!",
    "sentiment": "positive",
    "subreddit": "LangChain",
    "event_date": "2026-10",
    "post_flair": [
      "News"
    ],
    "post_title": "Introducing Rerank 5 (and a new feature) - available in LangChain",
    "prominence": "core",
    "source_url": "https://www.reddit.com/r/LangChain/comments/1xbprpo/introducing_rerank_5_and_a_new_feature_available/",
    "entity_role": "subject",
    "post_author": "Cohere",
    "upvote_ratio": 1,
    "mention_count": 5,
    "mention_surge": true,
    "subreddit_url": "https://www.reddit.com/r/LangChain/",
    "total_upvotes": 19,
    "comments_total": 13,
    "total_comments": 13,
    "event_date_text": "Today",
    "other_companies": [
      {
        "name": "LangChain",
        "role": "partner",
        "domain": "langchain.com"
      }
    ],
    "post_author_url": "https://www.reddit.com/user/Cohere/",
    "signal_category": "vendor_announcement",
    "comments_included": 6,
    "products_mentioned": [
      "Rerank 5",
      "Rerank 5 Pro",
      "Rerank 5 Fast",
      "Context Focus"
    ]
  }
}

curl https://signals.autobound.ai/v1/signals/dad5eb41-7c7b-5cc0-adfc-c0f1df3f3b40 \
  -H "X-API-KEY: $AUTOBOUND_API_KEY"

curl -X POST https://signals.autobound.ai/v1/companies/enrich \
  -H "X-API-KEY: $AUTOBOUND_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"domain":"cohere.com","limit":20}'