Visual-Language Understanding: leaderboard
Scale's SEAL Leaderboard evaluates top models' visual-language understanding, testing perception, logic, calculation, and common sense.
Metric: Score. Source: scale.com. Status: saturation imminent. 63 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Pro (Preview 03-25) | 54.65 |
| 2 | Gemini 2.5 Pro (Preview 06-05) | 54.63 |
| 3 | GPT-5.4 Pro (xHigh) | 53.89 |
| 4 | GPT-5 Pro | 52.39 |
| 5 | O4 Mini (High) | 51.79 |
| 6 | O4 Mini (Medium) | 51.66 |
| 7 | O3 Pro (High) | 51.63 |
| 8 | Gemini 3 Pro (Preview) | 51.49 |
| 9 | GPT-5.4 (xHigh) | 50.89 |
| 10 | Gemini 2.5 Pro (Preview 05-06) | 50.78 |
| 11 | GPT-5 Mini | 50.39 |
| 12 | O3 (High) | 50.07 |
| 13 | GPT-5 | 49.69 |
| 14 | O3 (Medium) | 49.59 |
| 15 | Gemini 2.5 Flash (Preview 05-20) | 49.15 |
Interactive version: theaggregate.ai/benchmark?slug=visual-language-understanding · How It Works · Data refreshed daily, snapshot 2026-09-05.