Nace AI has launched Drex for Search, extending its Drex decision models into document search capabilities specifically designed for AI agents. According to reporting by MarkTechPost, the new system achieved a 92.05 percent score on the BrowseComp-Plus benchmark using an average of 12.6 search calls, alongside a reduction in required agent turns on OfficeQA Pro.
Understanding Agentic Search Benchmarks
Benchmarks such as BrowseComp-Plus and OfficeQA Pro are designed to evaluate how efficiently artificial intelligence agents can navigate documents, retrieve accurate information, and complete multi-step tasks. In automated workflows, minimizing the number of search calls and agent turns typically correlates with lower computational latency and more reliable task execution.
What Remains Unclear
Despite the release, several operational details have not been disclosed. The announcement did not specify hardware requirements, pricing structures, or whether Drex for Search will be available as an open-source offering or a proprietary service. Furthermore, technical documentation detailing the exact integration process for existing agent architectures was absent from the initial report.
Developers and researchers will need to wait for further updates from Nace AI to determine the full practical impact of these decision models on production workloads.
Lead image: Lee Campbell / Pexels
