| Benchmark | Model | Rank | Results |
|---|---|---|---|
| arithmetic-reasoning-on-gsm8k | PretrainZero (Qwen3-30B-A3B-MoE-Base) | – | Accuracy: 94.4 |
| mathematical-reasoning-on-aime24 | PretrainZero (Qwen3-8B-Base) | – | Acc: 20.0 |
| mathematical-reasoning-on-math-500 | PretrainZero (Qwen3-8B-Base) | – | Accuracy: 81.9 |
| mathematical-reasoning-on-minerva-math | PretrainZero (Qwen3-30B-A3B-MoE-Base) | – | Accuracy: 41.2 |
| multi-task-language-understanding-on-mmlu-pro | PretrainZero (Qwen3-30B-A3B-MoE-Base) | – | Accuracy: 64.59 |