Humanity's Last Exam (Self-Reported, With 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), With Tools configuration — each lab's own agentic scaffold, so scores are not directly comparable across labs.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Around January 2027. 33 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V4.1 Flash | 63.9 | #36 |
| 2 | GLM-5.3 | 62.5 | #39 |
| 3 | Muse Spark Open | 61 | #54 |
| 4 | Ornith-1.5-397B | 56.1 | |
| 5 | Kimi K3 | 56 | #30 |
| 6 | Hy4 preview | 55.4 | #38 |
| 7 | GLM-5.3 Flash | 55.3 | #65 |
| 8 | GLM-5.2 | 54.7 | #73 |
| 9 | Kimi K2.6 | 54 | #104 |
| 10 | Hy3 | 53.2 | #93 |
| 11 | GLM-5.1 | 52.3 | #97 |
| 12 | Beam 1.0 | 51.1 | #91 |
| 13 | GLM-5 | 50.4 | #141 |
| 14 | Kimi K2.5 | 50.2 | #144 |
| 15 | Qwen 3.5 27B | 48.5 | #187 |
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-with-tools · How It Works · Data refreshed daily, snapshot 2026-10-20.