MobilePA-Bench - Skill Usage: leaderboard
Joint success when a mobile planning agent invokes pre-packaged composite skills instead of planning each step, pooled over the SOR and MTSR settings of MobilePA-Bench (MobilePA-Bench, arXiv 2608.23035v2, Table 3). Percent; higher is better.
Metric: Joint success (%; SOR and MTSR). Source: arxiv.org. Saturation forecast: Around February 2027. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 | 78 | #16 |
| 2 | Kimi K3 | 76.5 | #35 |
| 3 | Qwen 3.8 Max | 76.25 | #34 |
| 4 | Claude Fable 5 | 70.25 | #20 |
| 5 | GPT-5.6 Sol | 70 | #22 |
| 6 | GLM-5.2 | 67.75 | #80 |
| 7 | Claude Opus 4.8 | 67.5 | #39 |
| 8 | GPT-5.5 | 67.25 | #33 |
| 9 | Gemini 3.1 Pro (Preview) | 67 | #64 |
| 10 | Seed 2.1 Pro | 63.75 | #73 |
| 11 | Gemini 3.6 Flash | 63.5 | #57 |
| 12 | Qwen 3.7 Max | 53.75 | #85 |
| 13 | Kimi K2.6 | 46.5 | #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-skill-usage · How It Works · Data refreshed daily, snapshot 2026-11-05.