STEMVerse - Mathematics - Computational Mathematics: leaderboard
Metric: Few-shot accuracy (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 16 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen3-8B-Base | 59.15 |
| 2 | Qwen3-14B-Base | 56.34 |
| 3 | Qwen 2.5 3B Instruct | 55.63 |
| 4 | Qwen 2.5 14B | 53.52 |
| 5 | Qwen 2.5 7B | 45.07 |
| 6 | Qwen3-4B-Base | 44.37 |
| 7 | Qwen 2.5 3B | 43.66 |
| 8 | Qwen 2.5 14B Instruct | 39.44 |
| 9 | Qwen 2.5 7B Instruct | 38.03 |
| 10 | Llama 3.2 3B Instruct | 32.39 |
| 11 | Llama 3.1 8B Instruct | 30.99 |
| 12 | Llama 3.1 8B | 21.13 |
| 13 | Llama 3.2 3B | 6.34 |
Interactive version: theaggregate.ai/benchmark?slug=stemverse-mathematics-computational-mathematics · How It Works · Data refreshed daily, snapshot 2026-09-25.