AA Humanity's Last Exam — leaderboard
Artificial Analysis independent evaluation of HLE: 2,500 expert-vetted frontier questions across math, sciences, and humanities.
Metric: Accuracy (%). Source: artificialanalysis.ai. Status: saturation imminent. 540 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.6 Sol (Max) | 47.17 |
| 2 | Claude Opus 4.8 (Adaptive Reasoning, Max Effort) | 45.74 |
| 3 | Gemini 3.1 Pro (Preview) | 44.72 |
| 4 | GPT-5.6 Sol (xHigh) | 44.72 |
| 5 | Kimi K3 | 44.35 |
| 6 | GPT-5.5 (xHigh) | 44.3 |
| 7 | GPT-5.6 Sol (High) | 44.07 |
| 8 | GPT-5.5 (High) | 43 |
| 9 | GPT-5.6 Terra (Max) | 41.8 |
| 10 | GPT-5.4 (xHigh) | 41.61 |
| 11 | Gemini 3.5 Flash (High) | 40.96 |
| 12 | GPT-5.5 (Medium) | 40.59 |
| 13 | Grok 4.5 (High) | 40.27 |
| 14 | GLM-5.2 (Max) | 40.13 |
| 15 | GPT-5.6 Terra (xHigh) | 40.04 |
Interactive version: theaggregate.ai/benchmark?slug=aa-humanity-s-last-exam · How the rankings work · Data refreshed daily, snapshot 2026-07-22.