SuperCLUE-CPIF - Overall: leaderboard
SuperCLUE-CPIF evaluates Chinese precise instruction following across varied constraint types. Headline board of the newest workbook the superclueai.com vertical lists; score 0-100, higher is better.
Metric: Score. Source: www.superclueai.com. Saturation forecast: Estimated already saturated. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.1 (High) | 83.59 | #145 (GPT-5.1) |
| 2 | Gemini 3 Pro (Preview) | 83.08 | #71 |
| 3 | GPT-5 (High) | 82.56 | #105 (GPT-5) |
| 4 | DeepSeek V3.2 Exp (Thinking) | 74.36 | #241 (DeepSeek V3.2 Exp) |
| 5 | Claude Sonnet 4.5 (Thinking) | 72.82 | #148 (Claude Sonnet 4.5) |
| 6 | Kimi K2 (Thinking) | 67.18 | #252 (Kimi K2) |
| 7 | Hunyuan-T1-20250822 | 66.15 | #306 |
| 8 | Gemini 2.5 Pro | 63.59 | #159 |
| 9 | GLM-4.6 | 60.82 | #265 |
| 10 | MiniMax-M2 | 54.36 | #327 |
| 11 | Grok 4 | 53.85 | #189 |
| 12 | Qwen 3 Max | 51.79 | #213 |
| 13 | Doubao-Seed-1.6-thinking-250715 | 46.67 | #169 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=superclue-cpif-overall · How It Works · Data refreshed daily, snapshot 2026-11-05.