A recent study has demonstrated that leading artificial intelligence models can now pass the notoriously difficult Chartered Financial Analyst (CFA) Level III exam. Researchers from the AI-powered advisory firm Goodfin and the NYU Stern School of Business found that multiple large language models (LLMs) achieved scores well above the passing threshold, completing the test in just minutes.
Key Takeaways
- A joint study by Goodfin and NYU Stern tested 23 AI models against mock CFA Level III exams.
- OpenAI's o4-mini model achieved the highest score at 79.1%, surpassing the 65% passing grade.
- This marks a significant advancement, as previous AI models struggled with the open-ended essay questions of the Level III exam.
- Financial experts express caution, noting that AI lacks crucial human skills but may automate entry-level tasks essential for training junior analysts.
AI Demonstrates Advanced Financial Reasoning
The study, titled "Advanced Financial Reasoning at Scale," evaluated the capabilities of 23 prominent large language models. The results showed a clear leap in AI's ability to handle complex financial analysis and problem-solving, tasks that were previously exclusive to highly trained human professionals.
Among the top performers, OpenAI's o4-mini model led the group with a score of 79.1%. Other models also comfortably passed, including Google's Gemini 2.5 Flash at 77.3% and Anthropic's Claude Opus 4 at 74.9%. The success of these models is particularly noteworthy because the CFA Level III exam includes complex, open-ended essay questions requiring nuanced understanding and synthesis of information.
Top AI Performers on the Exam
Several AI models scored significantly above the CFA's estimated 65% passing score, showcasing their advanced capabilities.
- OpenAI o4-mini: 79.1%
- Google Gemini 2.5 Flash: 77.3%
- Anthropic Claude Opus 4: 74.9%
- xAI Grok-3: 73.8%
- DeepSeek DeepSeek-RI: 67.4%
A New Milestone for Artificial Intelligence
This achievement represents a rapid evolution in AI technology. Just last year, AI models were capable of passing the first two levels of the CFA exam, which are primarily multiple-choice. However, the unstructured, essay-based format of Level III remained a significant hurdle.
The ability to succeed on this final level suggests that AI can now engage in the kind of sophisticated reasoning required for expert-level roles in portfolio management, asset valuation, and wealth planning. According to Anna Joo Fee, CEO of Goodfin, the AI models accomplished this without any specific training tailored for the exam, indicating a broad and generalizable reasoning ability.
What is the CFA Charter?
The Chartered Financial Analyst (CFA) designation is a globally recognized credential in the investment management and financial analysis fields. Earning the charter requires passing three sequential, six-hour exams. The curriculum covers a wide range of topics, including ethical and professional standards, quantitative methods, economics, financial reporting, and portfolio management. The exam series is known for its low pass rates and is widely considered one of the most rigorous tests in the finance industry.
Industry Reactions and Future Implications
While the study's results are impressive, financial industry professionals have offered a measured response. They acknowledge the power of AI as a tool but emphasize the continued importance of human judgment and interpersonal skills in finance.
"If all elementary schools gave kids calculators to learn times tables, then when they get to algebra class, they’re not going to know what the heck they’re doing," said Michael DeMassa, founder of Forza Wealth Management. "AI is like a calculator for CFAs."
DeMassa expressed concern not about AI replacing seasoned analysts, but about its potential to automate the foundational tasks that junior analysts perform. He argued that these entry-level responsibilities are crucial for developing the core skills and deep understanding necessary for a successful career in finance. If AI handles all the preliminary work, the next generation of financial experts may lack the hands-on experience needed to tackle more complex challenges.
The Role of Human Expertise
Goodfin's CEO, Anna Joo Fee, noted that AI still has significant limitations. She pointed out that models cannot perform tasks that require uniquely human abilities, such as interpreting a client's body language or building trust through personal interaction. These skills are often essential in wealth management and financial advising.
The consensus among many experts is that AI will likely become an indispensable assistant for financial professionals rather than a replacement. It can rapidly process vast amounts of data, perform complex calculations, and generate initial analyses, freeing up human analysts to focus on higher-level strategy, client relationships, and qualitative judgments.
The trend of using AI for financial decisions is already growing. An eToro survey revealed that 13% of global retail investors report using tools like ChatGPT to help them pick stocks, highlighting a shift in how financial information is accessed and utilized.
The Broader Economic Context
The advancement of AI in finance is occurring alongside other significant economic trends. For instance, data indicates that foreigners' allocation to U.S. assets reached a record high of 32% in the second quarter, a level not seen since records began in 1968. This influx of international capital underscores the global interconnectedness of financial markets, where AI-driven tools are becoming increasingly prevalent.
As companies like Nvidia and Alibaba ramp up AI spending, and with major investments such as Nvidia's planned $100 billion in OpenAI, the technological infrastructure supporting these advanced models is expanding rapidly. This growth is expected to accelerate AI's integration into finance and many other sectors, fundamentally changing how analysis and decision-making are conducted.