AI models now capable of passing toughest CFA exam, study shows

Artificial intelligence has reached a milestone in professional finance. According to a CNBC report, a new study conducted by research led by Shilpi Nayak, Co-Founder & CTO of Goodfin and NYU Stern School of Business, has shown that advanced AI models can now pass the Chartered Financial Analyst (CFA) Level III exam, often regarded as one of the hardest tests in finance.

For humans, the CFA journey typically requires over 1,000 hours of study across several years. But researchers found that leading large language models (LLMs), including o4-mini, Gemini 2.5 Pro and Claude Opus. successfully cleared Level III mock exams in minutes using chain-of-thought prompting, a reasoning technique.

Rapid leap in AI capability

Previous studies had demonstrated that AI could manage Levels I and II of the CFA exam but failed at Level III, particularly because of the essay-style questions that demand nuanced reasoning. However, the latest results indicate that LLMs are now advanced enough to handle “specialised, high-stakes analytical reasoning required for professional financial decision-making.”

The CFA Level III exam is primarily centred on portfolio management and wealth planning, areas that require a deeper analytical approach.

Industry voices

“This research shows how quickly the technology is evolving and where it could take the industry,” said Anna Joo Fee, founder and CEO of GoodFin. While Fee acknowledged the breakthrough, she also stressed that AI would not replace human CFAs any time soon.

“There are things like context and intent that are hard for the machine to assess right now,” she told CNBC. “That’s where a human shines, in understanding body language and cues.”