#ModelAccuracyPaper
1QwQ-32B-Preview95.22link
2GPT-OSS (120B)94.25link
3DeepSeek-r1-Dist.91.28link
4Skywork-o191.28link
5Ours (+SFT)88.20link
6STAIR87.60link
7Self-Critique86.20link
8Shallow-Align85.80link
9Original84.50link
10DPO84.15link
11DeepSeek-R1 (70B)82.95link
12Gemma4 (31B)81.35link
13LLaMA-o179.38link
14SFT72.02link
GSM8k gsm8k Leaderboard