ADeptS-Bench - Attack Success (Desktop): leaderboard
ADeptS-Bench (2026) tests computer-use agents on single decision points in mobile and desktop screenshots. This board is the share of malicious desktop safety-stream tasks, with a threat embedded in the interface, in which the agent performs the unsafe action, in percent; lower is better.
Metric: Attack success rate (%, lower is better; share of the 501 malicious desktop tasks in which the model performs the unsafe action; paired benign and malicious GUI tasks with visual threats (phishing overlays, misleading buttons, prompt injection), graded on the next action against ground-truth boxes; mean of 10 runs at temperature 0.1). Source: arxiv.org. Saturation forecast: Around February 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.7 | 27.2 | #79 |
| 2 | Gemini 3.1 Pro (Preview) | 29.8 | #94 |
| 3 | GPT-5.4 | 36.3 | #123 |
| 4 | Qwen 3 VL 235B A22B | 76.3 | #479 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=adepts-bench-attack-success-desktop · How It Works · Data refreshed daily, snapshot 2026-11-13.