QMFOLBench - Mathematics: leaderboard
Metric: Accuracy (%; 720 mathematics-topic tasks). Source: arxiv.org. Saturation forecast: Estimated already saturated. 8 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) (High) | 98.75 |
| 2 | GPT-5.4 (High) | 97.08 |
| 3 | Qwen 3.5 27B (Thinking) | 93.06 |
| 4 | Claude Sonnet 4.6 (Thinking) | 79.03 |
| 5 | DeepSeek V3.2 (Thinking) | 77.92 |
| 6 | DeepSeek V3.2 (Non-reasoning) | 67.5 |
| 7 | GPT-5.4 (Non-reasoning) | 62.64 |
| 8 | Qwen 3 32B (Thinking) | 60.14 |
Interactive version: theaggregate.ai/benchmark?slug=qmfolbench-mathematics · How It Works · Data refreshed daily, snapshot 2026-09-25.