Reasoning Models Ace the CFA Exams (CFA mock-exam evaluation)

Size

980 questions across 8 mock exams (3 Level I, 2 Level II, 3 Level III)

Public split

none (official CFA Institute practice packs and paywalled AnalystPrep mocks)

Grading

MCQ accuracy. Constructed responses are graded by o4-mini against the AnalystPrep rubric (points out of 132 per exam). Results are reported zero-shot and with chain-of-thought.

Licence

Not stated

Submission

none

Products allowed

No

Latest version

arXiv v1 (2025-12-09)

What it does not cover

Exam knowledge only. No tools, real filings, live data or client advice.

Each score shows who reported it: the benchmark's authors (quoted from their paper), the benchmark's official board, or Moneda (run with the benchmark's own grader and rules). We do not copy other aggregators.

1. 97.6Gemini 3.0 Pro97.6
2. 96.1 GPT-596.1
3. 95.7Gemini 2.5 Pro95.7
4. 94.8 Grok 494.8
5. 94.6Claude Opus 4.194.6
6. 90.9DeepSeek-V3.190.9

1. 1Gemini 3.0 Pro97.697.6
2. 2 GPT-596.196.1
3. 3Gemini 2.5 Pro95.795.7
4. 4 Grok 494.894.8
5. 5Claude Opus 4.194.694.6
6. 6DeepSeek-V3.190.990.9

Reasoning Models Ace the CFA Exams (CFA mock-exam evaluation) leaderboard (Level I accuracy %, higher is better)

Rank	Model	Provider	Level I accuracy %	Setting	Source
1	Gemini 3.0 Pro	Google	97.6	zero-shot	Paper Table 5 Reported by the authors, reported 9 Dec 2025 · checked 1 Nov 2026
2	GPT-5	OpenAI	96.1	zero-shot	Paper Table 5 Reported by the authors, reported 9 Dec 2025 · checked 1 Nov 2026
3	Gemini 2.5 Pro	Google	95.7	zero-shot	Paper Table 5 Reported by the authors, reported 9 Dec 2025 · checked 1 Nov 2026
4	Grok 4	xAI	94.8	zero-shot	Paper Table 5 Reported by the authors, reported 9 Dec 2025 · checked 1 Nov 2026
5	Claude Opus 4.1	Anthropic	94.6	zero-shot	Paper Table 5 Reported by the authors, reported 9 Dec 2025 · checked 1 Nov 2026
6	DeepSeek-V3.1	DeepSeek	90.9	zero-shot	Paper Table 5 Reported by the authors, reported 9 Dec 2025 · checked 1 Nov 2026

1. 86.6Gemini 3.0 Pro86.6
2. 78.2Gemini 2.5 Pro78.2
3. 73.4Claude Opus 4.173.4
4. 71.8 GPT-571.8
5. 71.2 Grok 471.2
6. 70.8DeepSeek-V3.170.8

1. 1Gemini 3.0 Pro86.686.6
2. 2Gemini 2.5 Pro78.278.2
3. 3Claude Opus 4.173.473.4
4. 4 GPT-571.871.8
5. 5 Grok 471.271.2
6. 6DeepSeek-V3.170.870.8

Reasoning Models Ace the CFA Exams (CFA mock-exam evaluation) leaderboard (Level III constructed-response score %, higher is better)

Rank	Model	Provider	Level III constructed-response score %	Setting	Source
1	Gemini 3.0 Pro	Google	86.6	zero-shot, o4-mini grader	Paper Table 5 Reported by the authors, reported 9 Dec 2025 · checked 1 Nov 2026
2	Gemini 2.5 Pro	Google	78.2	zero-shot, o4-mini grader	Paper Table 5 Reported by the authors, reported 9 Dec 2025 · checked 1 Nov 2026
3	Claude Opus 4.1	Anthropic	73.4	zero-shot, o4-mini grader	Paper Table 5 Reported by the authors, reported 9 Dec 2025 · checked 1 Nov 2026
4	GPT-5	OpenAI	71.8	zero-shot, o4-mini grader	Paper Table 5 Reported by the authors, reported 9 Dec 2025 · checked 1 Nov 2026
5	Grok 4	xAI	71.2	zero-shot, o4-mini grader	Paper Table 5 Reported by the authors, reported 9 Dec 2025 · checked 1 Nov 2026
6	DeepSeek-V3.1	DeepSeek	70.8	zero-shot, o4-mini grader	Paper Table 5 Reported by the authors, reported 9 Dec 2025 · checked 1 Nov 2026

• Paper

Last checked 1 Nov 2026. Something out of date or wrong? Send a correction.