Mathematical Reasoning on MATH 500 (Accuracy and Classification Metrics)
98.5AccuracyDeepSeek-V3
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| DeepSeek-V3Model=DeepSeek-V32026.04 | 98.5 | 56.7 | 1.4 | 28.7 | 97.1 | 99.7 | |
| Qwen3Model=Qwen32026.04 | 98 | 51.6 | 0 | 17.5 | 92.1 | 96.9 | |
| Llama4-MaverickModel=Llama4-Maverick2026.04 | 90.3 | 83.3 | 77.9 | 81.2 | 80.3 | 93.6 | |
| Grok-4Model=Grok-42026.04 | 78.7 | 76.5 | 51.5 | 73.3 | 85.4 | 95.7 | |
| Tulu3.1-8BModel=Tulu3.1-8B2026.04 | 48.8 | 78.2 | 60.6 | 87.1 | 74.3 | 90.9 | |
| Llama3.1-8BModel=Llama3.1-8B2026.04 | 46.5 | 76.4 | 61.8 | 80.3 | 75.6 | 87.7 | |
| OLMo2-7BModel=OLMo2-7B2026.04 | 32.9 | 76.5 | 71 | 90.7 | 62.2 | 82.1 |