Google just released an autonomous research agent that scored 85.9% on BrowseComp, the benchmark for...
TL;DR · AI 摘要
Google 发布自主研究代理工具 Deep Research Max，基于 Gemini 3.1 Pro，BrowseComp 基准测试得分 85.9%，显著超越 GPT-5.4 和 Claude Opus。
核心要点
- Deep Research Max 在 BrowseComp 测试中得分为 85.9%，远超 GPT-5.4（58.9%）和 Claude Opus（45.1%）。
- 该工具支持夜间运行生成详尽报告，单次使用成本约 $2-$5。
- FactSet、S&P 和 PitchBook 正构建 MCP 服务器以整合其金融数据。
GPT-5.4 scored 58.9%. Claude Opus 4.6 scored 45.1%.
Deep Research Max is powered by Gemini 3.1 Pro. Runs overnight. Roughly $2–$5 per report. https://t.co/cxNfEZ3unj" / X
The Rundown AI on X: "Google just released an autonomous research agent that scored 85.9% on BrowseComp, the benchmark for locating hard-to-find facts online. GPT-5.4 scored 58.9%. Claude Opus 4.6 scored 45.1%. Deep Research Max is powered by Gemini 3.1 Pro. Runs overnight. Roughly $2–$5 per report. https://t.co/cxNfEZ3unj" / X
Don’t miss what’s happening
Google just released an autonomous research agent that scored 85.9% on BrowseComp, the benchmark for locating hard-to-find facts online. GPT-5.4 scored 58.9%. Claude Opus 4.6 scored 45.1%. Deep Research Max is powered by Gemini 3.1 Pro. Runs overnight. Roughly $2–$5 per report. FactSet, S&P, and PitchBook are all building MCP servers so it can plug directly into their financial data. Google, in their launch blog: "a nightly cron job triggering the generation of exhaustive due diligence reports for an analyst team by morning."
·
7
5
24
16
Read 7 replies