MobileWorldSafety - Task Completion Rate: leaderboard
MobileWorldSafety (Shanghai AI Laboratory, 2026) also tracks benign task utility under environmental injection attacks on real Android apps. The task completion rate is the share of scorable tasks in which the agent completes the benign task while resisting the attack, in percent, higher is better.
Metric: Task completion rate under attack (%; share of scorable tasks in which the agent completes the benign task while resisting the injection; 142 real Android-app tasks). Source: arxiv.org. Saturation forecast: Not forecast. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 397B A17B | 46.8 | #242 |
| 2 | Claude Sonnet 4.5 | 44.2 | #244 |
| 3 | Kimi K2.5 | 43.2 | #252 |
| 4 | Gemini 3 Pro | 30 | #135 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=mobileworldsafety-task-completion-rate · How It Works · Data refreshed daily, snapshot 2026-11-12.