ATLAS Finance benchmark: best of 11 frontier models scores 12% vs humans' 100%
garrytan · x · 2026-09-16
Garrett Lord unveiled ATLAS Finance, a benchmark giving 11 frontier models 100 expert-level finance tasks that take humans 15–30 hours each. Agents must navigate data rooms, email, chat, calendars, docs and Excel while handling dozens of persona-driven ambiguities and real-time updates. Result: humans score 100%, the best AI just 12% — evidence that autonomous knowledge work in banking, advisory and PE is nowhere near junior-professional standards. "If it's 95% right, it's 100% wrong."
More from AGI Musings
- Palmer Luckey asks: can you use datacenter RAM for things prohibited by the operator's terms of service? — avaitopiper · 2026-11-02
- On Terence Tao's blog, Frank Calegari calls on mathematicians to adapt after OpenAI's 700-paper release — Hesamation · 2026-11-02
- Polsia's Ben argues AI can autonomously run any human economic activity, including 80% of a company — LinusEkenstam · 2026-11-02
- AI Assistants Remember You, But Nothing Helps You Find What They Told You — Ok_Negotiation_2587 · 2026-11-02
- OpenAI and Anthropic claim AI-discovered proofs disprove the Exponential Time Hypothesis — emmanuelvivier · 2026-11-02
- Researcher: using a $20 AI subscription to fix my phone hints AI may suppress GDP — stanislavfort · 2026-11-02