--- title: Nace AI Releases Drex for Search for AI Agents: 92.05%... description: Nace AI has released Drex for Search, a retrieval layer that lets AI agents find cited evidence in documents instead of relying on top-k chunk retrieval. It... canonical: https://daily.dev/posts/nace-ai-releases-drex-for-search-for-ai-agents-92-05-on-browsecomp-plus-with-12-6-search-calls-lk0lva9rn twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Nace AI Releases Drex for Search for AI Agents: 92.05% on BrowseComp-Plus With 12.6 Search Calls | daily.dev og:description: Nace AI has released Drex for Search, a retrieval layer that lets AI agents find cited evidence in documents instead of relying on top-k chunk retrieval. It... og:url: https://daily.dev/posts/nace-ai-releases-drex-for-search-for-ai-agents-92-05-on-browsecomp-plus-with-12-6-search-calls-lk0lva9rn og:image: https://api.daily.dev/og/posts/LK0lVa9rN.png og:image:alt: Nace AI Releases Drex for Search for AI Agents: 92.05% on BrowseComp-Plus With 12.6 Search Calls og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Nace AI Releases Drex for Search for AI Agents: 92.05% on BrowseComp-Plus With 12.6 Search Calls **[Machine Learning News](https://daily.dev/sources/mlnews)** · 6 min read · 0 upvotes · 0 comments ## Summary Nace AI has released Drex for Search, a retrieval layer that lets AI agents find cited evidence in documents instead of relying on top-k chunk retrieval. It scores 92.05% on BrowseComp-Plus with GPT-5, using 12.61 search calls per question, 16.6 times fewer than the only entry ranked above it. On VisDocAgentBench it hits 77.5% R@1 with an open-weight Qwen 27B planner, versus 67.5% for Claude Opus 5. On OfficeQA Pro, across 9 agent and model pairs, it cut agent turns and tokens while improving accuracy. It's accessible via CLI, Python and TypeScript SDKs, MCP, and REST. This article is sponsored by Nace AI. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What accuracy does Drex for Search achieve on the BrowseComp-Plus benchmark? Drex for Search reports 92.05% accuracy on BrowseComp-Plus using GPT-5 as the answering model, ranking #2 of 86 entries on the public leaderboard. It uses only 12.61 search calls per question, 16.6 times fewer than the top-ranked entry (AI21 Labs' hybrid Qwen2-7B plus Reason-ModernColBERT system), which needed 209.27 calls to reach 95.18%. _Teams weighing agentic retrieval accuracy against search-call cost can compare benchmark tradeoffs like this on daily.dev._ ### How does Drex for Search compare to Claude Opus 5 on document retrieval benchmarks? On VisDocAgentBench, where an agent must find one target page among 2,375 arXiv pages, Drex for Search scores 77.5% R@1 using an open-weight Qwen 27B planner, 10 points above Claude Opus 5's 67.5%. Recall drops with multi-hop difficulty: 97.5% on direct questions, 77.5% with one bridge, and 57.5% with two bridges. _Developers choosing a document retrieval agent can track comparisons like this on daily.dev before committing._ ### How much does Drex for Search reduce the number of agent steps needed for document Q&A tasks? On the OfficeQA Pro benchmark of 133 questions over U.S. Treasury Bulletins, adding Drex for Search to 9 different agent and model pairs cut agent turns by 62% and tokens by 56% on average, while improving accuracy by 6% and running 2.3 times faster. For example, Gemini CLI with Gemini 3.1 Pro dropped from 22.9 to 10.4 turns while accuracy rose from 54.9% to 61.7%. _Anyone optimizing agent step counts and token costs for document search can follow benchmarks like this on daily.dev._ ## Similar posts on daily.dev - [Codex is now the recommended agent in JetBrains IDEs](https://daily.dev/posts/codex-is-now-the-recommended-agent-in-jetbrains-ides-q3gxvk24g) · JetBrains · 1 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#mcp](https://daily.dev/tags/mcp), [#rag](https://daily.dev/tags/rag) [View this post on daily.dev](https://daily.dev/posts/nace-ai-releases-drex-for-search-for-ai-agents-92-05-on-browsecomp-plus-with-12-6-search-calls-lk0lva9rn) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Nace AI Releases Drex for Search for AI Agents: 92.05% on BrowseComp-Plus With 12.6 Search Calls","url":"https://daily.dev/posts/nace-ai-releases-drex-for-search-for-ai-agents-92-05-on-browsecomp-plus-with-12-6-search-calls-lk0lva9rn","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/nace-ai-releases-drex-for-search-for-ai-agents-92-05-on-browsecomp-plus-with-12-6-search-calls-lk0lva9rn"},"datePublished":"2026-10-26T16:51:26.427Z","dateModified":"2026-10-27T23:26:16.908Z","description":"Nace AI has released Drex for Search, a retrieval layer that lets AI agents find cited evidence in documents instead of relying on top-k chunk retrieval. It...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/5fbb9f4590b77d634a465e44f7520987?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/5fbb9f4590b77d634a465e44f7520987?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Machine Learning News","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Machine Learning News","logo":"https://media.daily.dev/image/upload/s--0PoEtdkd--/f_auto/v1750409401/squads/marktechpost?_a=BAMClqZW0","url":"https://daily.dev/squads/mlnews"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/nace-ai-releases-drex-for-search-for-ai-agents-92-05-on-browsecomp-plus-with-12-6-search-calls-lk0lva9rn","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,mcp,rag","timeRequired":"PT6M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Machine Learning News","item":"https://daily.dev/squads/mlnews"},{"@type":"ListItem","position":3,"name":"Nace AI Releases Drex for Search for AI Agents: 92.05% on BrowseComp-Plus With 12.6 Search Calls"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/nace-ai-releases-drex-for-search-for-ai-agents-92-05-on-browsecomp-plus-with-12-6-search-calls-lk0lva9rn#faq","mainEntity":[{"@type":"Question","name":"What accuracy does Drex for Search achieve on the BrowseComp-Plus benchmark?","acceptedAnswer":{"@type":"Answer","text":"Drex for Search reports 92.05% accuracy on BrowseComp-Plus using GPT-5 as the answering model, ranking #2 of 86 entries on the public leaderboard. It uses only 12.61 search calls per question, 16.6 times fewer than the top-ranked entry (AI21 Labs' hybrid Qwen2-7B plus Reason-ModernColBERT system), which needed 209.27 calls to reach 95.18%. Teams weighing agentic retrieval accuracy against search-call cost can compare benchmark tradeoffs like this on daily.dev."}},{"@type":"Question","name":"How does Drex for Search compare to Claude Opus 5 on document retrieval benchmarks?","acceptedAnswer":{"@type":"Answer","text":"On VisDocAgentBench, where an agent must find one target page among 2,375 arXiv pages, Drex for Search scores 77.5% R@1 using an open-weight Qwen 27B planner, 10 points above Claude Opus 5's 67.5%. Recall drops with multi-hop difficulty: 97.5% on direct questions, 77.5% with one bridge, and 57.5% with two bridges. Developers choosing a document retrieval agent can track comparisons like this on daily.dev before committing."}},{"@type":"Question","name":"How much does Drex for Search reduce the number of agent steps needed for document Q\u0026A tasks?","acceptedAnswer":{"@type":"Answer","text":"On the OfficeQA Pro benchmark of 133 questions over U.S. Treasury Bulletins, adding Drex for Search to 9 different agent and model pairs cut agent turns by 62% and tokens by 56% on average, while improving accuracy by 6% and running 2.3 times faster. For example, Gemini CLI with Gemini 3.1 Pro dropped from 22.9 to 10.4 turns while accuracy rose from 54.9% to 61.7%. Anyone optimizing agent step counts and token costs for document search can follow benchmarks like this on daily.dev."}}]} ```