AndroidDaily - Up to 2 Constraints: leaderboard

AndroidDaily (2026), a benchmark of everyday tasks for mobile GUI agents in real closed-source Android apps, restricted to tasks with at most two constraints. Sessions are judged against task-specific guidelines; the score is the task success rate in percent; higher is better.

Metric: Task success rate (%) on tasks with at most two constraints, 350 tasks in 94 real closed-source Android apps, end-to-end sessions judged by the GRADE evaluator (Gemini 3 Pro backbone, 87.37 percent agreement with human review) against three-tier task guidelines; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 12 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Pro74.2#145
2Gemini 3 Flash69.4#175
3Seed 1.858.9#252
4Seed 2.0 Pro55.6#222
5UI-TARS-1.554
6GLM-4.6V43.5#710
7Step-GUI-8B42.7
8Step-GUI-30A341.9
9GUI-Owl-1.5-32B-Think33.6
10UI-Venus-1.5-8B31.5
11GUI-Owl-1.5-8B-Think29.5
12MAI-UI-8B15.3

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=androiddaily-up-to-2-constraints · How It Works · Data refreshed daily, snapshot 2026-11-21.