In just a few minutes, AI easily passed the CFA Level III exam.
```
Latest research shows that multiple AI models can now pass the prestigious CFA Level 3 exam within minutes, while humans usually require several years and around 1,000 hours of study to complete it.
Researchers from NYU Stern School of Business and AI wealth management platform GoodFin tested 23 large language models and found that cutting-edge reasoning models such as o4-mini, Gemini 2.5 Pro, and Claude Opus were able to successfully pass the CFA Level 3 simulated exam.
These models use "chain-of-thought prompting" techniques, which effectively addressed essay question types that previously troubled AI. GoodFin founder and CEO Anna Joo Fee stated:
I believe this technology definitely has the potential to reshape the entire industry in the future.
AI Models Achieve Major Breakthrough in CFA Level 3 Exam Barriers
Research from two years ago found that AI models could pass the CFA Level 1 and 2 exams, but faced significant challenges with essay sections of the Level 3 exam.
The latest research confirms that the rapid development of AI technology has already overcome this key bottleneck.
The research team conducted comprehensive tests on 23 large language models, with particular attention paid to whether the AI models possessed the "specialist, high-risk analytical reasoning abilities required for professional financial decisions" in order to assess their ability to handle multiple-choice and essay questions in the CFA Level 3 simulated exam.
The research results showed that when dealing with complex financial problems, reasoning models performed significantly better than traditional non-reasoning models. Especially, cutting-edge reasoning models using "chain-of-thought prompting" successfully passed the relevant exam tests.
Among them, Gemini 2.5 Pro scored the highest with 3.44 points in essay questions, and also ranked first overall (multiple-choice and essay) with 2.1 points.
It is noteworthy that the domestically developed KIMI K2 model performed best in multiple-choice questions, with an accuracy rate as high as 78.3%, surpassing Google’s Gemini 2.5 Pro and GPT-5.
The research used three prompting strategies: zero-shot, self-consistency, and self-discovery, with the self-consistency strategy achieving the best performance score at 73.4%.
In the cost-effectiveness analysis, Llama 3.1 8B Instant achieved the best cost efficiency score of 5468, while Palmyra Fin had the fastest average response time of 0.3 seconds.
Although AI performs well in standardized exams, industry experts believe there are still limitations to completely replacing human financial professionals. Fee emphasized:
In areas such as situational understanding and intent judgment, machines still find it difficult to make accurate assessments. This is where humans have an advantage, being able to understand clients’ body language and various cues.
Risk Warning and DisclaimerThe market has risks, investment needs caution. This article does not constitute personal investment advice and does not take into account the individual user's specific investment objectives, financial situation, or needs. Users should consider whether any opinions, views, or conclusions in this article are appropriate to their particular circumstances. Investment is at your own risk. ```