Prim - Generation: leaderboard
Metric: Generation accuracy (%): exact match of the model's own zero-shot chain-of-thought final answer on 182 HLE-Verified mathematics problems; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.4 (High) | 67.03 |
| 2 | Qwen 3.6 27B | 52.75 |
| 3 | GPT-5.4 Mini (High) | 50.55 |
| 4 | Qwen 3.5 27B | 47.8 |
| 5 | GPT-5.4 Nano (High) | 44.51 |
| 6 | GPT-OSS-20B (High) | 43.41 |
| 7 | Qwen 3.5 9B | 38.46 |
| 8 | Qwen 3.5 4B | 26.37 |
| 9 | DeepSeek R1 Distill Qwen 32B | 18.13 |
| 10 | DeepSeek R1 0528 Qwen3 8B | 17.03 |
| 11 | DeepSeek R1 Distill Qwen 14B | 15.93 |
| 12 | DeepSeek-R1-Distill-Qwen-7B | 13.19 |
Interactive version: theaggregate.ai/benchmark?slug=prim-generation · How It Works · Data refreshed daily, snapshot 2026-10-04.