UXBench (Mobile) - Badge Mismatch: leaderboard
UXBench (2026, mobile UI) tests whether multimodal models can diagnose user-experience defects from real mobile app screenshots with two- or three-option questions balanced between defective and normal screens. This board is accuracy (0-100) on questions about badge content that does not match the landing page; unparseable answers count as wrong; higher is better.
Metric: Accuracy (0-100) on badge content inconsistent with the landing page (trustworthiness): two- or three-option multiple-choice UX-defect diagnosis questions on real mobile app screenshots (2,000 questions over eight tasks, balanced positive and negative cases), temperature 0, 8,192-token output limit, a response that cannot be parsed counts as wrong; printed as a 0-1 accuracy and x100 here; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 70 | #250 |
| 2 | Qwen 2.5 VL 72B Instruct | 69 | #846 |
| 3 | Claude Sonnet 4 | 68 | #400 |
| 4 | Qwen 3 VL 235B A22B Instruct | 62 | #522 |
| 5 | Claude 3.7 Sonnet | 61 | #458 |
| 6 | Qwen 3 VL 235B A22B (Thinking) | 54 | #451 |
| 7 | Qwen 3 VL 4B (Thinking) | 52 | #1296 |
| 8 | GLM-4.1V-9B (Thinking) | 32 | #1060 (GLM-4.1V-9B) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=uxbench-mobile-badge-mismatch · How It Works · Data refreshed daily, snapshot 2026-11-13.