Factual Knowledge Question Answering on MMLU-Pro
80Accuracy (Acc)Llama4-Maverick
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Llama4-MaverickAveraged over=three LLM raters2026.04 | 80 | 63.9 | 30.4 | 44 | 84.1 | 97.1 | |
| DeepSeek-V3Averaged over=three LLM raters2026.04 | 79.4 | 62 | 17.5 | 39.4 | 92.3 | 98.8 | |
| Grok-4Averaged over=three LLM raters2026.04 | 69.4 | 61.5 | 16.6 | 42.3 | 88.9 | 98.3 | |
| Qwen3Averaged over=three LLM raters2026.04 | 50.6 | 66.3 | 35.9 | 66.2 | 71.1 | 92.1 | |
| Llama3.1-8BAveraged over=three LLM raters2026.04 | 40.4 | 62.6 | 38.2 | 70.6 | 50.5 | 91.2 | |
| Tulu3.1-8BAveraged over=three LLM raters2026.04 | 38.4 | 70 | 51.4 | 86.2 | 53.2 | 89.1 | |
| OLMo2-7BAveraged over=three LLM raters2026.04 | 28.6 | 67.8 | 58.1 | 91.9 | 36.4 | 84.8 |