Humanity's Last Exam (Self-Reported, No Tools): leaderboard
Metric: Accuracy (%). Source: huggingface.co. 53 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Kimi K3 | 56 |
| 2 | Hy3 | 53.2 |
| 3 | Inkling | 46 |
| 4 | Qwen3.8 2.4T A95B | 43.6 |
| 5 | Hy4 preview | 43.4 |
| 6 | GLM-5.2 | 40.5 |
| 7 | DeepSeek V4 Pro | 37.7 |
| 8 | Qwen3.8-Flash-Next | 35.9 |
| 9 | DeepSeek V4 Flash | 34.8 |
| 10 | Kimi K2.6 | 34.7 |
| 11 | Inkling Small | 31.6 |
| 12 | GLM-5.1 | 31 |
| 13 | Qwen 3.8 27B | 30.8 |
| 14 | GLM-5 | 30.5 |
| 15 | Hy3-preview | 30 |
Interactive version: theaggregate.ai/benchmark?slug=humanity-s-last-exam-self-reported-no-tools · How It Works · Data refreshed daily, snapshot 2026-09-05.