Seeing Isn't Knowing: leaderboard
Seeing Isn't Knowing — do VLMs know when visual evidence is insufficient (occlusion, misleading perspective); self-reported occlusion-split accuracy.
Metric: Occlusion All (self-reported). Source: benchmarklist.com. Saturation forecast: Around April 2027. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash | 57.1 | #99 |
| 2 | Gemini 2.5 Flash | 53.2 | #245 |
| 3 | GPT-5 Mini | 49.9 | #186 |
| 4 | GPT-4o | 48.4 | #343 |
| 5 | GPT-5.4 | 48.1 | #79 |
| 6 | Qwen 2.5 VL 7B | 48 | #643 |
| 7 | InternVL3-8B | 47.5 | #620 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=seeing-isn-t-knowing · How It Works · Data refreshed daily, snapshot 2026-10-20.