Mathematical Reasoning on GSM8K (Acc, vbal, FP, TN, FN, TP)
97.8AccuracyQwen3
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Qwen3Model=Qwen32026.04 | 97.8 | 65.3 | 15.7 | 35.5 | 92.7 | 96.7 | |
| Llama4-MaverickModel=Llama4-Maverick2026.04 | 97.2 | 64 | 22.3 | 42.9 | 91.6 | 99.1 | |
| DeepSeek-V3Model=DeepSeek-V32026.04 | 95.9 | 56.9 | 7.3 | 26.3 | 95 | 99.1 | |
| Grok-4Model=Grok-42026.04 | 92.7 | 62.7 | 7.1 | 50 | 94.2 | 99.7 | |
| Tulu3.1-8BModel=Tulu3.1-8B2026.04 | 88.2 | 78.8 | 50.6 | 76.8 | 89.8 | 98.2 | |
| Llama3.1-8BModel=Llama3.1-8B2026.04 | 77.9 | 80.4 | 59.5 | 76.4 | 89.3 | 96.3 | |
| OLMo2-7BModel=OLMo2-7B2026.04 | 75.4 | 76 | 46.9 | 72.3 | 88.6 | 96.2 |