Humanity's Sixth Sense
Multimodal · 2026-10-07
Scale AI and Elorian's intuitive visual reasoning benchmark: a frozen release of 522 human-authored tasks (288 image and 234 silent-video tasks), across four domains and eleven subdomains. Pass@1 averages three attempts per task; Claude Opus 5 judges against reference answers and 723 rubric criteria. Published confidence intervals use a task-cluster bootstrap. Released October 7, 2026.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-6 Astra | 53.6 |
| GPT-6.1 Sol | 46.6 |
| Claude Opus 5.5 | 44.6 |
| Gemini 3.8 Flash | 41.6 |
| Claude Fable 5.1 | 40.8 |
| Gemini 3.7 Flash | 39.8 |
| Muse Spark 1.3 | 37.4 |
| Claude Fable 5 | 34.5 |
| Gemini 3.5 Flash | 32.6 |
| Gemini 3.6 Flash | 31.9 |
| GPT-6 Sol | 31.2 |
| Qwen3.8 Flash | 30.9 |
| Claude Opus 5 | 30.5 |
| GPT-5.6 Sol | 30.0 |
| GLM-5.3 Flash | 26.8 |
| Kimi K3 | 25.5 |
| MiMo-V2.6-Flash | 25.4 |
| MiMo-V2.6-Pro | 25.2 |
| Sonnet 5 | 24.8 |
| Muse Spark 1.2 | 24.5 |
| Opus 4.8 | 23.5 |
| GPT-5.6 Terra | 22.7 |
| GPT-5.6 Luna | 21.6 |
| GPT-6 Luna | 21.0 |