Humanity's Last Exam (Self-Reported, No Tools): leaderboard
PhD-level questions at the frontier of human expert knowledge. Lab-submitted scores from the official cais/hle dataset board on Hugging Face (model-card numbers, none independently verified), No Tools configuration; mostly open-weights models.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Around September 2027. 50 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Ornith-1.5-397B | 44.6 | |
| 2 | Qwen3.8 2.4T A95B | 43.6 | #55 |
| 3 | Hy4 preview | 43.4 | #38 |
| 4 | GLM-5.2 | 40.5 | #73 |
| 5 | DeepSeek V4 Pro | 37.7 | #102 |
| 6 | Ling 3.1 Flash | 37.44 | #71 |
| 7 | DeepSeek V4.1 Flash | 36.8 | #36 |
| 8 | Beam 1.0 | 36 | #91 |
| 9 | Qwen3.8-Flash-Next | 35.9 | #72 |
| 10 | DeepSeek V4 Flash | 34.8 | #137 |
| 11 | Kimi K2.6 | 34.7 | #104 |
| 12 | Inkling Small | 31.6 | #138 |
| 13 | GLM-5.1 | 31 | #97 |
| 14 | Qwen 3.8 27B | 30.8 | #105 |
| 15 | GLM-5 | 30.5 | #141 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=humanity-s-last-exam-self-reported-no-tools · How It Works · Data refreshed daily, snapshot 2026-10-20.