TMMLU+ - Humanities: leaderboard
Metric: Accuracy (%). Source: huggingface.co. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 5 | 94.6 |
| 2 | Claude Fable 5 | 90.02 |
| 3 | Gemini 3.1 Pro (Preview) (Thinking) | 84.91 |
| 4 | GPT-5.6 Sol | 82.3 |
| 5 | Claude Sonnet 5 | 81.72 |
| 6 | DeepSeek V4 Pro (0813) | 79.63 |
| 7 | DeepSeek V4 Flash | 77.42 |
| 8 | Kimi K3 | 77.31 |
| 9 | Qwen 3.7 Max | 76.03 |
| 10 | Hy3 | 74.75 |
| 11 | GPT-5.6 Terra | 72.49 |
| 12 | GPT-5.6 Luna | 70.81 |
| 13 | GLM-5.2 | 70.28 |
| 14 | Grok 4.3 | 67.67 |
| 15 | MiMo-V2.5 | 67.32 |
Interactive version: theaggregate.ai/benchmark?slug=tmmlu-plus-humanities · How It Works · Data refreshed daily, snapshot 2026-09-19.