Mathematical Reasoning on AIME 25 (Accuracy, Acc Delta UltraQuant)
90AccuracyBF16
Evaluation results
| Method | Links | ||
|---|---|---|---|
| BF16Model=Qwen3.5-A3B2026.06 | 90 | — | |
| BF16Model=MiniMax-M2.52026.06 | 86.67 | — | |
| TQ4/4Model=Qwen3.5-A3B2026.06 | 83.33 | — | |
| Ultra-TQModel=Qwen3.5-A3B2026.06 | 83.33 | — | |
| Ultra-TQModel=MiniMax-M2.52026.06 | 83.33 | — | |
| TQ4/4Model=MiniMax-M2.52026.06 | 80 | — | |
| UltraQuantModel=Qwen3.5-A3B2026.06 | 76.67 | 13.33 | |
| UltraQuantModel=MiniMax-M2.52026.06 | 76.67 | 10 | |
| BF16Model=Qwen2.5-72B2026.06 | 20 | — | |
| TQ4/4Model=Qwen2.5-72B2026.06 | 16.67 | — | |
| UltraQuantModel=Qwen2.5-72B2026.06 | 16.67 | 3.33 |