Science Question Answering on GPQA Diamond (Accuracy, Quantization Delta)
84.34AccuracyBF16
Evaluation results
| Method | Links | ||
|---|---|---|---|
| BF16Model=MiniMax-M2.52026.06 | 84.34 | — | |
| Ultra-TQModel=MiniMax-M2.52026.06 | 83.84 | — | |
| TQ4/4Model=MiniMax-M2.52026.06 | 83.33 | — | |
| UltraQuantModel=MiniMax-M2.52026.06 | 82.32 | 2.02 | |
| BF16Model=Qwen3.5-A3B2026.06 | 79.8 | — | |
| UltraQuantModel=Qwen3.5-A3B2026.06 | 79.8 | 0 | |
| TQ4/4Model=Qwen3.5-A3B2026.06 | 78.28 | — | |
| TQ4/4Model=Qwen2.5-72B2026.06 | 52.53 | — | |
| Ultra-TQModel=Qwen2.5-72B2026.06 | 52.53 | — | |
| UltraQuantModel=Qwen2.5-72B2026.06 | 51.01 | 1.52 | |
| BF16Model=Qwen2.5-72B2026.06 | 49.49 | — |