Prim - Execution: leaderboard
Prim's execution score: exact-match accuracy on 182 Humanity's Last Exam mathematics problems when the gold mathematical primitive (the solution's key idea) is supplied with the problem; higher is better (Prim, 2026).
Metric: Execution accuracy (%): exact-match accuracy on the 182 problems when the gold mathematical primitive is given with the problem; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 12 models tracked. Updated 4 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | GPT-5.4 | High | 86.81 | #138 |
| 2 | Qwen 3.6 27B | 78.57 | #352 | |
| 3 | Qwen 3.5 27B | 71.98 | #378 | |
| 4 | GPT-5.4 Mini | High | 71.98 | #405 |
| 5 | GPT-5.4 Nano | High | 68.68 | #651 |
| 6 | Qwen 3.5 9B | 61.54 | #829 | |
| 7 | GPT-OSS-20B | High | 60.99 | #1166 |
| 8 | Qwen 3.5 4B | 56.04 | #1235 | |
| 9 | DeepSeek R1 Distill Qwen 32B | 43.41 | #1813 | |
| 10 | DeepSeek R1 0528 Qwen3 8B | 39.56 | #2190 | |
| 11 | DeepSeek R1 Distill Qwen 14B | 39.01 | #2706 | |
| 12 | DeepSeek-R1-Distill-Qwen-7B | 32.42 | #7912 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=prim-execution · How It Works · Data refreshed daily, snapshot 2026-11-20.