Reasoning Models Ace the CFA Exams - Jaisal Patel, et al, arXiv

A new study just found that six leading AI models now pass all three levels of the Chartered Financial Analyst (CFA) certification exams, with Gemini 3.0 Pro scoring a record high of 97.6% on Level 1. Researchers tested GPT-5, Gemini 3.0 Pro, Claude Opus 4.1, Grok 4, and DeepSeek-V3.1 across 980 questions spanning all exam tiers. GPT-5 topped Level II at 94.3%, while Gemini 3.0 Pro dominated the most difficult constructed-response section with 92%. In 2023, GPT 3.5 failed the first two levels, and GPT-4 passed only Level I — with the leap to near-perfect scores taking roughly 24 months. An NYU study in September also showed models passing all three levels, but with scores in the 70s vs. the near-perfect numbers of current frontier systems. Why it matters: Acing a standardized test and handling daily demands of financial analysis are still very different things, but the speed of improvement on these exams is wild — and models mastering finance knowledge could shift the profession’s value toward human skills like client judgment and relationship management.

https://elinke1c.daily.therundown.ai/ss/c/u001.Q334NVcZU4O6L6VKRz8ijMNgampOZl8dF7lEf1g2D9OlYV1yRyWOM_tuvK4JPLtgxGixLXSHzjenudvEAMcncQ0fOaJRL5wOoJZ-nGDhb6iMK5fSSScQQLvOXuCEhin6jixPZE8vkXxQcH3XpS4CD4Vn080vcd7XACgoE6Ztk7oT5dWoX84lVf8D9xlWg5V1dHwwqib7Nqje-0R0NdQfb4rXe7W9T2Ik3op-Dk8Kzio5-5wEav7sVioMELH5DbbE/4mh/XPw-J0iFT52qovcivcc8GA/h10/h001.2_0S4gW5trZ4tx2u8n_KgFAQk0ejbfuW6JDtpH39_gs