SuperCLUE-SWE (December 2025) - Overall: leaderboard
SuperCLUE-SWE resolves real GitHub issues from Chinese open-source projects with Chinese problem descriptions. December 2025 edition (2025年12月): headline 总分 of the bug修复能力榜单 sheet. 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: Around April 2027. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 50 | #71 |
| 2 | Claude Opus 4.5 (20251101) | 47 | #81 |
| 3 | Gemini 3 Flash (Preview) | 45 | #90 |
| 4 | Claude Sonnet 4.5 | 40 | #148 |
| 5 | GLM-5 | 35 | #149 |
| 6 | GPT-5 (High) | 35 | #105 (GPT-5) |
| 7 | GPT-5.2 (High) | 28 | #115 (GPT-5.2) |
| 8 | GPT-5.1 | 27 | #145 |
| 9 | DeepSeek V3.2 | 25 | #210 |
| 10 | Kimi K2 (Thinking) | 24 | #252 (Kimi K2) |
| 11 | Qwen3 Coder | 20 | #304 |
| 12 | MiMo-V2-Flash | 19 | #378 |
| 13 | GLM-4.6 | 18 | #265 |
| 14 | DeepSeek V3.2 Exp (Thinking) | 18 | #241 (DeepSeek V3.2 Exp) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=superclue-swe-december-2025-overall · How It Works · Data refreshed daily, snapshot 2026-11-05.