QuantEval - Knowledge QA (CoT): leaderboard
QuantEval (Peking University and collaborators, 2026) knowledge questions: 660 multiple-choice and fill-in-the-blank items on quantitative-finance concepts, terminology and theory, drawn from textbooks, papers, regulatory filings and glossaries and validated by domain experts. Accuracy in percent, higher is better, with chain-of-thought prompting and greedy decoding; an LLM grader extracts each final answer and checks it against the reference. GPT-4o and Claude Sonnet 4.5 helped generate and validate items.
Metric: Accuracy (%; chain-of-thought prompting). Source: arxiv.org. Saturation forecast: Around February 2027. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 86 | #244 |
| 2 | GPT-5 | 83.9 | #163 |
| 3 | Qwen 3 14B | 75.9 | #1235 |
| 4 | Gemini 2.5 Pro | 66.7 | #262 |
| 5 | DeepSeek R1 Distill Qwen 14B | 66.7 | #2594 |
| 6 | Qwen 3 8B | 65.5 | #1838 |
| 7 | Qwen 3 4B | 64.4 | #2297 |
| 8 | DeepSeek R1 Distill Llama 8B | 50.6 | #6773 |
| 9 | Qwen 3 30B A3B | 48.5 | #1124 |
| 10 | DeepSeek-R1-Distill-Qwen-7B | 42.5 | #8045 |
| 11 | DeepSeek R1 Distill Qwen 1.5B | 37.2 | #9257 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=quanteval-knowledge-qa-cot · How It Works · Data refreshed daily, snapshot 2026-11-12.