MMLU-Pro
knowledgeA harder, reasoning-heavy rebuild of MMLU with 10 answer choices instead of 4, cutting saturation and prompt sensitivity across 14 disciplines.
Official benchmark pageModel rankings on MMLU-Pro
| # | Model | Score | As of | Source |
|---|---|---|---|---|
| 1 |  Gemini 3.1 Pro Preview | 92.6% | Feb 19, 2026 | cite |
| 2 |  DeepSeek V4 Pro | 87.5% | Apr 24, 2026 | cite |
| 3 | DeepSeek V4 Flash | 86.4% | Apr 24, 2026 | cite |
| 4 |  Kimi K2 Thinking | 84.6% | Nov 6, 2025 | cite |
| 5 |  Llama 4 Maverick | 80.5% | Apr 5, 2025 | cite |
Scores are self-reported or from primary evaluations, each linked to its source. Test conditions (tools, shots, prompt) vary between labs — see the source for details.