8 months ago
AI Models Pass CFA Exams
Recently, some very smart computer programs called AI models took a really hard test for finance jobs called the CFA exam.
These tests are known to be very difficult, and they check if someone knows a lot about money, math, and how to make good decisions.
The AI models did really well, with one called Gemini 3.0 Pro getting a super high score of 97.6% on the first level.
Another model, GPT-5, did great on the second level.
The researchers found that these new AI models are much better at thinking through problems step by step and understanding the context, which older models struggled with.
However, the AI models still have some trouble with questions about ethics and writing detailed answers.
This means that while AI is getting really good at finance tests, it can't replace human finance professionals just yet.
It's a big deal because it shows that AI is getting smarter and can help with important finance jobs, but it also makes people think about what makes human finance experts special.
Six leading AI models passed all three levels of the CFA exams, considered the 'Gold Standard' for investment professionals.
Gemini 3.0 Pro scored 97.6% on Level I, while GPT-5 topped Level II with 94.3%.
Newer AI models excel in multi-step reasoning and context tracking, unlike older models that struggled with nuance and application.
AI models still face challenges in ethics and constructed-response sections, indicating room for improvement.
Passing CFA exams by AI does not mean they can replace human analysts, but it shows AI's growing capability in structured financial reasoning.
- Who
- Researchers tested AI models on CFA exams
- What
- AI models passed all three levels of CFA exams
- Where
- Not specified, likely global
- When
- Recent study, specific date not mentioned
- Why
- To assess AI's capability in financial reasoning and professional exams
Key facts
- Exam
- Chartered Financial Analyst (CFA) certification
- AI Models Tested
- Gemini 3.0 Pro, GPT-5, Claude Opus 4.1, Grok 4, DeepSeek-V3.1
- Highest Score
- Gemini 3.0 Pro with 97.6% on Level I
- Top Model in Level II
- GPT-5 with 94.3%
- Key Improvement in AI
- Better multi-step reasoning and context tracking
- Weakness in AI
- Ethics and constructed-response sections
- Implications
- AI can handle structured financial reasoning tasks
Quotes
Tech company that launched the Gemini 3 Flash AI model
“Gemini 3 Flash demonstrates that speed and scale don’t have to come at the cost of intelligence. It delivers frontier performance on PhD-level reasoning and knowledge benchmarks like GPQA Diamond (90.4%) and Humanity’s Last Exam (33.7% without tools), rivaling larger frontier models, and significantly outperforming even the best 2.5 model, Gemini 2.5 Pro, across a number of benchmarks.”
livemint.com
Sources
Related news
US Department of War Expands GenAI.mil With ChatGPT, Grok
AI Model Releases Outpace Independent Testing, Raising Evaluation Concerns
SpaceXAI unveils Grok 4.6, claims frontier-level performance at lower cost
Elon Musk's Grok Bot Beta Released on Desktop and iOS
Grok Bot Can See, Click and Type in Early Beta
