Vellum - Humanity's Last Exam — leaderboard
Vellum's independently-run Humanity's Last Exam evaluation. Expert-level questions across diverse academic domains.
Metric: Accuracy (%). Source: www.vellum.ai. Status: saturation imminent. 28 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Mythos 5 | 64.5 |
| 2 | Claude Opus 4.8 | 57.9 |
| 3 | Claude Sonnet 5 | 57.4 |
| 4 | GLM-5.2 | 54.7 |
| 5 | Kimi K2.6 | 54 |
| 6 | DeepSeek V4 Flash | 51.6 |
| 7 | DeepSeek V4 Pro | 48.2 |
| 8 | Gemini 3 Pro | 45.8 |
| 9 | Kimi K2 (Thinking) | 44.9 |
| 10 | Gemini 3.1 Pro (Preview) | 44.4 |
| 11 | GPT-5.5 Pro | 43.1 |
| 12 | GPT-5.5 | 41.4 |
| 13 | Gemini 3.5 Flash | 40.2 |
| 14 | Claude Opus 4.6 | 40 |
| 15 | GPT-5 | 35.2 |
Interactive version: theaggregate.ai/benchmark?slug=vellum-humanity-s-last-exam · How the rankings work · Data refreshed daily, snapshot 2026-07-22.