Nace AI's Drex for Search Tops Snowflake MADQA With 94.7% Accuracy
testingcatalog · x · 2026-10-28
Nace AI's Drex for Search has taken the top spot on Snowflake's MADQA leaderboard, answering 94.7% of 500 document-based questions correctly. It ships as an MCP tool that plugs into Claude Code, Codex, Cursor and OpenCode.
On the 133-question OfficeQA Pro set, Nace reports accuracy gains of 2-7 points alongside sharply fewer agent turns:
- Gemini CLI + Gemini 3.1 Pro: 54.9% → 61.7%, turns 22.9 → 10.4
- Codex + GPT-6 sol: 66.2% → 70.7%, turns 14.3 → 8.6
- Claude Code + Opus 5.5: 66.5% → 69.9%, turns 18.5 → 9.1
- Hermes Agent + Qwen3.8-27B: 56.4% → 60.2%, turns 20.1 → 6.5
- Exception: Codex + GPT-6 luna slipped 1.5 points to 53.4%
The vendor claims 6% more correct answers, 62% fewer steps, 56% fewer tokens and 2.3x faster completion. A demo also pulled a value out of a scanned page in Apple's 10-K, getting 7 of 8 questions right and missing only the scanned-page one.
Related event: Nace AI's Drex for Search Tops Snowflake MADQA Leaderboard(2 posts)→
More from coding & agent
- In the agentic world, the lines between engineering disciplines blur — yunta_tsai · 2026-10-28
- LLMs supply intelligence, agents do the work — do SaaS interfaces become obsolete? — ThomasBuildLab · 2026-10-28
- Codex scans your connectors/MCPs unprompted in the background, and you can't tell — kenziyuliu · 2026-10-28
- UCLA and Tsinghua paper probes how far coding agents can go in visual generation — ucla-cmllab · 2026-10-28
- 'Claude Duet has won': devs say use provider-agnostic harnesses as the top model keeps rotating — p_e_cooper · 2026-10-28
- ArkenOS deploys swarm of autonomous scientists; Sam Altman approves — _AustinCalvert_ · 2026-10-28