ActiveVision (Agents): leaderboard
ActiveVision solved by tool-using agent harnesses with image-manipulation tools, separately from the chain-of-thought track. Native correct/85 accuracy retains each Claude Code or Codex harness and its published reasoning effort.
Metric: Accuracy (%). Source: activevision.dev. Saturation forecast: Rough model projection: around 2026. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra (xHigh) (ActiveVision Run: %7B%22model%22%3A%22GPT-6%20Astra%20Codex%22%2C%22setting%22%3A%22xhigh%20%2B%20tools%22%2C%22track%22%3A%22ActiveVision%20%28Agents%29%22%7D) | 91.76 | |
| 2 | Opus 5.5 (xHigh) (ActiveVision Run: %7B%22model%22%3A%22Opus%205.5%20Claude%20Code%22%2C%22setting%22%3A%22xhigh%20%2B%20tools%22%2C%22track%22%3A%22ActiveVision%20%28Agents%29%22%7D) | 82.35 | |
| 3 | Sonnet 5.5 (xHigh) (ActiveVision Run: %7B%22model%22%3A%22Sonnet%205.5%20Claude%20Code%22%2C%22setting%22%3A%22xhigh%20%2B%20tools%22%2C%22track%22%3A%22ActiveVision%20%28Agents%29%22%7D) | 78.82 | |
| 4 | Fable 5.1 (xHigh) (ActiveVision Run: %7B%22model%22%3A%22Fable%205.1%20Claude%20Code%22%2C%22setting%22%3A%22xhigh%20%2B%20tools%22%2C%22track%22%3A%22ActiveVision%20%28Agents%29%22%7D) | 76.47 | |
| 5 | Opus 5 (xHigh) (ActiveVision Run: %7B%22model%22%3A%22Opus%205%20Claude%20Code%22%2C%22setting%22%3A%22xhigh%20%2B%20tools%22%2C%22track%22%3A%22ActiveVision%20%28Agents%29%22%7D) | 68.24 | |
| 6 | Fable 5 (xHigh) (ActiveVision Run: %7B%22model%22%3A%22Fable%205%20Claude%20Code%22%2C%22setting%22%3A%22xhigh%20%2B%20tools%22%2C%22track%22%3A%22ActiveVision%20%28Agents%29%22%7D) | 50.59 | |
| 7 | GPT-5.6 Sol (xHigh) (ActiveVision Run: %7B%22model%22%3A%22GPT-5.6%20Sol%20Codex%22%2C%22setting%22%3A%22xhigh%20%2B%20tools%22%2C%22track%22%3A%22ActiveVision%20%28Agents%29%22%7D) | 44.71 | |
| 8 | GPT-5.5 (xHigh) (ActiveVision Run: %7B%22model%22%3A%22GPT-5.5%20Codex%22%2C%22setting%22%3A%22xhigh%20%2B%20tools%22%2C%22track%22%3A%22ActiveVision%20%28Agents%29%22%7D) | 37.65 | |
| 9 | Opus 4.8 (xHigh) (ActiveVision Run: %7B%22model%22%3A%22Opus%204.8%20Claude%20Code%22%2C%22setting%22%3A%22xhigh%20%2B%20tools%22%2C%22track%22%3A%22ActiveVision%20%28Agents%29%22%7D) | 24.71 |
Interactive version: theaggregate.ai/benchmark?slug=activevision-agents · How It Works · Data refreshed daily, snapshot 2026-10-23.