Benchmarks · reasoning
Humanity's Sixth Sense
Scale AI and Elorian's intuitive visual reasoning benchmark: 522 image and silent-video tasks covering temporal, spatial, social and abstract inference. Pass@1 averaged over three attempts per task; all rubric criteria must pass. Published intervals are 95% task-cluster bootstrap intervals.
27 models · 27 results · data as of 2026-10-29 · labs.scale.com ↗
Leaderboard
Gemini 4 Argon high
All results
27 rows| # | Model | Effort | Humanity's Sixth Sense |
|---|---|---|---|
| 1 | Gemini 4 Argon | high | 55.4% ±4.2 |
| 2 | GPT-6 Astra | max | 53.6% ±4.1 |
| 3 | GPT-6.1 Sol | max | 46.6% ±3.8 |
| 4 | Opus 5.5 | xhigh | 44.6% ±3.8 |
| 5 | Gemini 3.8 Flash | high | 41.6% ±3.7 |
| 6 | Fable 5.1 | max | 40.8% ±3.6 |
| 7 | Gemini 3.7 Flash | high | 39.8% ±3.7 |
| 8 | Muse Spark 1.3 | max | 37.4% ±3.5 |
| 9 | Fable 5 | max | 34.5% ±3.7 |
| 10 | Qwen 3.8 Max | max | 33.0% ±3.6 |
| 11 | Gemini 3.5 Flash | high | 32.6% ±3.4 |
| 12 | Gemini 3.6 Flash | high | 31.9% ±3.4 |
| 13 | GPT-6 Sol | max | 31.2% ±3.4 |
| 14 | Qwen 3.8 Flash | max | 30.9% ±3.3 |
| 15 | Opus 5 | max | 30.5% ±3.5 |
| 16 | DeepSeek V4.1 Flash 1017 | max | 30.1% ±3.3 |
| 17 | GPT-5.6 Sol | max | 30.0% ±3.3 |
| 18 | GLM 5.3 Flash | max | 26.8% ±3.0 |
| 19 | Kimi K3 | max | 25.5% ±3.1 |
| 20 | MiMo V2.6 Flash | high | 25.4% ±2.9 |
| 21 | MiMo V2.6 Pro | high | 25.2% ±2.9 |
| 22 | Sonnet 5 | max | 24.8% ±3.2 |
| 23 | Muse Spark 1.2 | xhigh | 24.5% ±3.1 |
| 24 | Opus 4.8 | max | 23.5% ±3.2 |
| 25 | GPT-5.6 Terra | max | 22.7% ±3.2 |
| 26 | GPT-5.6 Luna | max | 21.6% ±3.2 |
| 27 | GPT-6 Luna | max | 21.0% ±3.0 |