SuperCLUE-Ondevice (September 2025) - Safety: leaderboard
SuperCLUE-OnDevice evaluates on-device phone-class models; the roster also scores several large cloud models as reference points. September 2025 edition (2025年9月): safety sheet total covering values, responsible answers and refusal of harmful content. Each dated workbook is a separate evaluation round, kept apart from the current board. Score 0-100, higher is better.
Metric: Score. Source: www.superclueai.com. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Doubao-Seed-1.6-thinking-250715 | 94 | #169 |
| 2 | Gemini 2.5 Pro | 90.67 | #159 |
| 3 | Qwen 2.5 VL 7B | 90.61 | #641 |
| 4 | GLM-4.5V | 90 | #358 |
| 5 | Gemma 3 4B | 80.22 | #1088 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=superclue-ondevice-september-2025-safety · How It Works · Data refreshed daily, snapshot 2026-11-05.