Nace AI Launches Drex for Search: 92.05% on BrowseComp-Plus with Just 12.6 Search Calls Per Question
By Asif Razzaq | October 26, 2026
Nace AI has released Drex for Search, extending its Drex decision models to document search. The product is a retrieval layer designed to help both people and AI agents locate evidence quickly and reliably. Coding agents such as Claude Code, Codex, Cursor, and OpenCode can call it through CLI, SDKs, MCP, or REST. In Nace's own evaluation, Drex for Search scores 92.05% on BrowseComp-Plus with GPT-5, using an average of 12.6 search calls per question.
TL;DR
- Scale: A hosted search system built to search over 1M+ files. It pairs with an answering model such as GPT-5, Gemini 3.5 Flash, or Qwen 27B.
- Deployment: Runs on Nace's managed cloud, a dedicated cloud, your own AWS or Azure account, or on-premises and air-gapped environments.
- Performance: Ranked #1 or #2 on three of four public leaderboards, according to Nace.
Why It Matters
As agentic AI workflows mature in 2026, retrieval quality has become the limiting factor for agent reliability. Agents that search too often waste tokens and latency; agents that search too little miss critical evidence. Drex for Search targets that trade-off directly: high answer accuracy at a notably low search-call count.
Availability
Nace AI offers Drex for Search across managed cloud, dedicated cloud, customer-owned AWS or Azure accounts, and on-premises or air-gapped deployments, giving enterprises flexible options for data residency and security requirements.
This is a developing story. Details on pricing, latency benchmarks, and independent third-party evaluations were not available at the time of publication.
via MarkTechPost