Prim - Execution: leaderboard
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 January 2027. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.4 (High) | 86.81 |
| 2 | Qwen 3.6 27B | 78.57 |
| 3 | Qwen 3.5 27B | 71.98 |
| 4 | GPT-5.4 Mini (High) | 71.98 |
| 5 | GPT-5.4 Nano (High) | 68.68 |
| 6 | Qwen 3.5 9B | 61.54 |
| 7 | GPT-OSS-20B (High) | 60.99 |
| 8 | Qwen 3.5 4B | 56.04 |
| 9 | DeepSeek R1 Distill Qwen 32B | 43.41 |
| 10 | DeepSeek R1 0528 Qwen3 8B | 39.56 |
| 11 | DeepSeek R1 Distill Qwen 14B | 39.01 |
| 12 | DeepSeek-R1-Distill-Qwen-7B | 32.42 |
Interactive version: theaggregate.ai/benchmark?slug=prim-execution · How It Works · Data refreshed daily, snapshot 2026-10-04.