Humanity's Last Exam
General QA · 2025-01-24
A set of 2,500 expert-authored questions spanning over 100 academic subjects, designed to test the limits of frontier AI models on problems that require deep, specialized knowledge.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5.1 | 60.9 |
| Claude Fable 5 | 59.0 |
| Claude Mythos Preview | 56.8 |
| Claude Opus 5 | 56.4 |
| Gemini 4 Argon | 55.6 |
| GPT-6 Astra | 54.8 |
| GPT-6.1 Sol | 51.3 |
| Kimi K3.1 Pro | 51.1 |
| Fugu Ultra v1.0 | 50.0 |
| Opus 4.8 | 49.8 |
| Muse Spark 1.1 | 49.4 |
| Gemini 3 Deep Think | 48.4 |
| Muse Wonder | 47.6 |
| Opus 4.7 | 46.9 |
| Step 5 Preview | 46.5 |
| Gemini 3.8 Flash | 46.2 |
| Claude Sonnet 5.5 | 46.1 |
| Gemini 3.7 Flash | 45.6 |
| GPT-5.6 Sol | 45.5 |
| Muse Spark Open | 44.8 |
| Gemini 3.1 Pro Preview | 44.4 |
| GPT-5.4 Pro | 44.3 |
| Sonnet 5 | 44.1 |
| GPT-5.5 | 43.8 |
| Kimi K3 | 43.5 |
| Muse Spark 1.3 | 43.2 |
| GPT-5.5 Pro | 43.1 |
| DeepSeek V4 Pro 0813 | 42.7 |
| DeepSeek V4.1 Flash 1017 | 42.6 |
| GLM-5.2 | 41.4 |
| Muse Spark | 40.6 |
| GPT-5.4 | 40.3 |
| Opus 4.6 | 40.0 |
| Grok 4.6 | 39.7 |
| Grok 4.5 | 39.3 |
| GLM-5.3 | 39.1 |
| DeepSeek V4.1 Flash | 36.8 |
| Gemini 3 Flash Preview | 36.6 |
| GPT-5.2 Pro | 36.6 |
| Qwen3.8-Flash-Next | 35.9 |