ScaleWoB - Tasks Without Return Values: leaderboard
The 74 ScaleWoB tasks that are pure UI operations, checked programmatically on the final app state. ScaleWoB (Liu et al., Tsinghua University and Xiaomi, 2026) runs mobile GUI agents on 120 manually verified tasks in 63 web-built replicas of real apps that coding agents synthesized; agents see only screenshots, have at most 100 steps, and results are averaged over two runs. Success rate in percent; higher is better.
Metric: Success rate (%; programmatic check of the final app state; 74 UI-operation tasks, mean of two runs). Source: arxiv.org. Saturation forecast: Around March 2027. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 45.27 | #111 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=scalewob-tasks-without-return-values · How It Works · Data refreshed daily, snapshot 2026-11-12.