MobilePA-Bench - Overall: leaderboard
Weighted overall success of mobile planning agents in an executable sandbox with live app databases across 13 domains and 212 tools: 50% basic tool use, 10% sub-agent collaboration, 20% memory usage and 20% skill usage (MobilePA-Bench, arXiv 2608.23035v2, Table 3). Percent; higher is better.
Metric: Weighted success (%; 50/10/20/20). Source: arxiv.org. Saturation forecast: Around February 2027. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 | 75.52 | #16 |
| 2 | Claude Fable 5 | 75.31 | #20 |
| 3 | Kimi K3 | 73.01 | #35 |
| 4 | Qwen 3.8 Max | 72.51 | #34 |
| 5 | Gemini 3.6 Flash | 71.21 | #57 |
| 6 | Gemini 3.1 Pro (Preview) | 71.18 | #64 |
| 7 | GLM-5.2 | 67.71 | #80 |
| 8 | Claude Opus 4.8 | 65.52 | #39 |
| 9 | Qwen 3.7 Max | 64.71 | #85 |
| 10 | Seed 2.1 Pro | 63.65 | #73 |
| 11 | GPT-5.6 Sol | 62.68 | #22 |
| 12 | GPT-5.5 | 61.44 | #33 |
| 13 | Kimi K2.6 | 55.63 | #113 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mobilepa-bench-overall · How It Works · Data refreshed daily, snapshot 2026-11-05.