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

Mock CFA exams at all three levels: multiple choice at Levels I-III, plus constructed-response essays at Level III. Results are judged against CFA pass thresholds.

## Reasoning Models Ace the CFA Exams (CFA mock-exam evaluation): Level I accuracy % (higher is better, 6 models)

| Rank | Model | Maker | Score | Metric | Setting | Reported by | Source | URL | Date |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| 1 | Gemini 3.0 Pro | Google | 97.6 | Level I accuracy % | zero-shot | Reported by the authors | Paper Table 5 | https://arxiv.org/html/2512.08270v1 | 9 Dec 2025 |
| 2 | gpt-5 | OpenAI | 96.1 | Level I accuracy % | zero-shot | Reported by the authors | Paper Table 5 | https://arxiv.org/html/2512.08270v1 | 9 Dec 2025 |
| 3 | Gemini 2.5 Pro | Google | 95.7 | Level I accuracy % | zero-shot | Reported by the authors | Paper Table 5 | https://arxiv.org/html/2512.08270v1 | 9 Dec 2025 |
| 4 | grok-4 | xAI | 94.8 | Level I accuracy % | zero-shot | Reported by the authors | Paper Table 5 | https://arxiv.org/html/2512.08270v1 | 9 Dec 2025 |
| 5 | Claude Opus 4.1 | Anthropic | 94.6 | Level I accuracy % | zero-shot | Reported by the authors | Paper Table 5 | https://arxiv.org/html/2512.08270v1 | 9 Dec 2025 |
| 6 | DeepSeek V3.1 | DeepSeek | 90.9 | Level I accuracy % | zero-shot | Reported by the authors | Paper Table 5 | https://arxiv.org/html/2512.08270v1 | 9 Dec 2025 |

## Reasoning Models Ace the CFA Exams (CFA mock-exam evaluation): Level III constructed-response score % (higher is better, 6 models)

| Rank | Model | Maker | Score | Metric | Setting | Reported by | Source | URL | Date |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| 1 | Gemini 3.0 Pro | Google | 86.6 | Level III constructed-response score % | zero-shot, o4-mini grader | Reported by the authors | Paper Table 5 | https://arxiv.org/html/2512.08270v1 | 9 Dec 2025 |
| 2 | Gemini 2.5 Pro | Google | 78.2 | Level III constructed-response score % | zero-shot, o4-mini grader | Reported by the authors | Paper Table 5 | https://arxiv.org/html/2512.08270v1 | 9 Dec 2025 |
| 3 | Claude Opus 4.1 | Anthropic | 73.4 | Level III constructed-response score % | zero-shot, o4-mini grader | Reported by the authors | Paper Table 5 | https://arxiv.org/html/2512.08270v1 | 9 Dec 2025 |
| 4 | gpt-5 | OpenAI | 71.8 | Level III constructed-response score % | zero-shot, o4-mini grader | Reported by the authors | Paper Table 5 | https://arxiv.org/html/2512.08270v1 | 9 Dec 2025 |
| 5 | grok-4 | xAI | 71.2 | Level III constructed-response score % | zero-shot, o4-mini grader | Reported by the authors | Paper Table 5 | https://arxiv.org/html/2512.08270v1 | 9 Dec 2025 |
| 6 | DeepSeek V3.1 | DeepSeek | 70.8 | Level III constructed-response score % | zero-shot, o4-mini grader | Reported by the authors | Paper Table 5 | https://arxiv.org/html/2512.08270v1 | 9 Dec 2025 |

Scores are never combined across benchmarks. Each row names its metric, source and date.

## About this benchmark

- Maker: SecureFinAI Lab, Columbia University (with RPI and UNC Chapel Hill)
- Year: 2025
- Use case: Finance knowledge
- 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.
- Last checked: 1 Nov 2026
- Links: paper https://arxiv.org/abs/2512.08270
- Directory page: https://benchmarks.finance/benchmarks/cfa-reasoning-models-ace/
- Leaderboard: https://benchmarks.finance/leaderboards/cfa-reasoning-models-ace/

Source: https://benchmarks.finance/leaderboards/cfa-reasoning-models-ace/

Created and maintained by Moneda.