CAIS Vision Capabilities Index: leaderboard
BenchmarkList computes this index as the plain mean of six CAIS AI Dashboard measures, for models that have all six: EnigmaEval puzzles, IntPhys2 physics, ERQA embodied reasoning, MindCube spatial navigation, SpatialViz mental visualisation and ART. ART is the dashboard's Agent Red Teaming attack success rate, where lower is better, so this mean, ranked higher is better, mixes directions. Source: dashboard.safe.ai.
Metric: Vision Capabilities Index (imported). Source: dashboard.safe.ai. Source: benchmarklist.com. Saturation forecast: Around January 2028. 43 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra | 71.5 | #6 |
| 2 | Claude Opus 5.5 | 68.1 | #3 |
| 3 | GPT-6.1 Sol | 68 | #7 |
| 4 | Gemini 3.5 Flash | 65.9 | #60 |
| 5 | Gemini 3 Flash | 64.9 | #99 |
| 6 | Claude Opus 5 | 63.9 | #12 |
| 7 | Gemini 3.7 Flash | 63.7 | #23 |
| 8 | Kimi K3 | 63.2 | #30 |
| 9 | Gemini 3.8 Flash | 63.2 | #21 |
| 10 | Gemini 3.1 Pro (Preview) | 63.1 | #58 |
| 11 | Claude Fable 5.1 | 63.1 | #9 |
| 12 | Claude Fable 5 | 62.1 | #16 |
| 13 | GPT-5.6 Sol | 62 | #19 |
| 14 | Kimi K2.6 | 61.2 | #104 |
| 15 | Grok 4.5 | 60.8 | #48 |
No result here: #8 Gemini 4 Argon, #10 Muse Wonder Full.
Interactive version: theaggregate.ai/benchmark?slug=cais-vision-capabilities-index · How It Works · Data refreshed daily, snapshot 2026-10-20.