ActiveVision: leaderboard
activevision.dev "exam for active observers": iterative visual reasoning where models scan, traverse and transform images over multiple steps. Native correct/85 accuracy retains each reasoning effort and the explicitly labeled text-only ablation as separate configurations.
Metric: Accuracy (%). Source: activevision.dev. Saturation forecast: Rough model projection: around 2026. 58 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | gpt-6-astra (xHigh) (ActiveVision Run: %7B%22model%22%3A%22gpt-6-astra%22%2C%22setting%22%3A%22xhigh%22%2C%22track%22%3A%22ActiveVision%22%7D) | 76.47 | |
| 2 | gpt-6.1-sol (Max) (ActiveVision Run: %7B%22model%22%3A%22gpt-6.1-sol%22%2C%22setting%22%3A%22max%22%2C%22track%22%3A%22ActiveVision%22%7D) | 71.76 | |
| 3 | gpt-6-astra (Max) (ActiveVision Run: %7B%22model%22%3A%22gpt-6-astra%22%2C%22setting%22%3A%22max%22%2C%22track%22%3A%22ActiveVision%22%7D) | 70.59 | |
| 4 | gpt-6-astra (High) (ActiveVision Run: %7B%22model%22%3A%22gpt-6-astra%22%2C%22setting%22%3A%22high%22%2C%22track%22%3A%22ActiveVision%22%7D) | 69.41 | |
| 5 | gpt-6.1-sol (xHigh) (ActiveVision Run: %7B%22model%22%3A%22gpt-6.1-sol%22%2C%22setting%22%3A%22xhigh%22%2C%22track%22%3A%22ActiveVision%22%7D) | 68.24 | |
| 6 | gpt-6-astra (Medium) (ActiveVision Run: %7B%22model%22%3A%22gpt-6-astra%22%2C%22setting%22%3A%22medium%22%2C%22track%22%3A%22ActiveVision%22%7D) | 65.88 | |
| 7 | gpt-6.1-sol (High) (ActiveVision Run: %7B%22model%22%3A%22gpt-6.1-sol%22%2C%22setting%22%3A%22high%22%2C%22track%22%3A%22ActiveVision%22%7D) | 64.71 | |
| 8 | gpt-6.1-sol (Medium) (ActiveVision Run: %7B%22model%22%3A%22gpt-6.1-sol%22%2C%22setting%22%3A%22medium%22%2C%22track%22%3A%22ActiveVision%22%7D) | 58.82 | |
| 9 | gpt-6-astra (Low) (ActiveVision Run: %7B%22model%22%3A%22gpt-6-astra%22%2C%22setting%22%3A%22low%22%2C%22track%22%3A%22ActiveVision%22%7D) | 52.94 | |
| 10 | gpt-6.1-sol (Low) (ActiveVision Run: %7B%22model%22%3A%22gpt-6.1-sol%22%2C%22setting%22%3A%22low%22%2C%22track%22%3A%22ActiveVision%22%7D) | 51.76 |
Interactive version: theaggregate.ai/benchmark?slug=activevision · How It Works · Data refreshed daily, snapshot 2026-10-24.