SuperCLUE-Writing
Models write original Chinese prose.
SuperCLUE reports scores in a public workbook.
- Made by
- SuperCLUE
- Kind
- Creative writing
- Level
- University
- Scored
- Graded by people
- Answers
- Open-ended
- Measures
- Capability
- Subject
- Languages and literature
- Language
- Chinese
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- A host
- Strongest model tested
- GPT-5.5 #21 on the capability index
- Last model added
- May 2026
- Scores refreshed
- November 20th, 2026
Claude Opus 4.7 leads with 88.0 among models released by May 2026 Claude Opus 4.7 leads with 88.0 among models released by May 2026
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.
The newest model with a score here came out in May 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 1 Claude Opus 4.7 max / API 88.0
- 2
ERNIE 5.1 API 87.6
- 3
Gemini 3.1 Pro API 86.7
- 4 Kimi K2.6 API 86.7
- 5 Doubao-Seed-2.0-Pro-260215 API 86.5
- 6
Qwen 3.6 Max (Preview) API 86.5
- 7 Grok 4.20 reasoning, API 86.2
- 8 GLM-5.1 API 85.7
- 9 GPT-5.5 API 85.2
- 10
DeepSeek-V4-Pro max / API 80.9
- 11 MiMo-V2-Pro API 78.9
- 12
DeepSeek-V4-Flash max / API 76.6
- 13 HY3-Preview(think_high) API 75.3
- 14
MiniMax-M2.7 API 56.5
Source: Official SuperCLUE-Writing leaderboard
- 1 Claude Opus 4.7 max / API 88.0
- 2
ERNIE 5.1 API 87.6
- 3
Gemini 3.1 Pro API 86.7
- 4 Kimi K2.6 API 86.7
- 5
Qwen 3.6 Max (Preview) API 86.5
- 6 Grok 4.20 reasoning, API 86.2
- 7 GLM-5.1 API 85.7
- 8 GPT-5.5 API 85.2
- 9
DeepSeek-V4-Pro max / API 80.9
- 10 MiMo-V2-Pro API 78.9
- 11
DeepSeek-V4-Flash max / API 76.6
- 12
MiniMax-M2.7 API 56.5
Source: Official SuperCLUE-Writing leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Opus 4.7 | 88.0 | max / API | Apr 2026 | Official leaderboard |
| 2 | | 87.6 | API | May 2026 | Official leaderboard |
| 3 | | 86.7 | API | Feb 2026 | Official leaderboard |
| 4 | Kimi K2.6 | 86.7 | API | Apr 2026 | Official leaderboard |
| 5 | Doubao-Seed-2.0-Pro-260215 | 86.5 | API | Official leaderboard | |
| 6 | | 86.5 | API | Apr 2026 | Official leaderboard |
| 7 | Grok 4.20 | 86.2 | reasoning, API | Feb 2026 | Official leaderboard |
| 8 | GLM-5.1 | 85.7 | API | Apr 2026 | Official leaderboard |
| 9 | GPT-5.5 | 85.2 | API | Apr 2026 | Official leaderboard |
| 10 | | 80.9 | max / API | Apr 2026 | Official leaderboard |
| 11 | MiMo-V2-Pro | 78.9 | API | Mar 2026 | Official leaderboard |
| 12 | | 76.6 | max / API | Apr 2026 | Official leaderboard |
| 13 | HY3-Preview(think_high) | 75.3 | API | Official leaderboard | |
| 14 | | 56.5 | API | Mar 2026 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Opus 4.7 | 88.0 | max / API | Apr 2026 | Official leaderboard |
| 2 | | 87.6 | API | May 2026 | Official leaderboard |
| 3 | | 86.7 | API | Feb 2026 | Official leaderboard |
| 4 | Kimi K2.6 | 86.7 | API | Apr 2026 | Official leaderboard |
| 5 | | 86.5 | API | Apr 2026 | Official leaderboard |
| 6 | Grok 4.20 | 86.2 | reasoning, API | Feb 2026 | Official leaderboard |
| 7 | GLM-5.1 | 85.7 | API | Apr 2026 | Official leaderboard |
| 8 | GPT-5.5 | 85.2 | API | Apr 2026 | Official leaderboard |
| 9 | | 80.9 | max / API | Apr 2026 | Official leaderboard |
| 10 | MiMo-V2-Pro | 78.9 | API | Mar 2026 | Official leaderboard |
| 11 | | 76.6 | max / API | Apr 2026 | Official leaderboard |
| 12 | | 56.5 | API | Mar 2026 | Official leaderboard |
Rankings by part
We read 26 parts from the benchmark’s own leaderboard in 5 groups: 内容质量, 内容风格, 内容创意, 主流题材创作 and 约束创作. Each ranks models on one part of the test. GPT-5.5 leads 6 of them, and 6 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
内容质量
12 models we track have a score here.
- 1 Claude Opus 4.7 90.8
- 2
ERNIE 5.1 89.6
- 3
Qwen 3.6 Max (Preview) ↑2 86.7
- 4 GLM-5.1 ↑3 86.1
- 5 Grok 4.20 ↑1 86.1
- 6
Gemini 3.1 Pro ↓3 85.0
- 7 Kimi K2.6 ↓3 84.1
- 8 GPT-5.5 81.3
- 9
DeepSeek-V4-Pro 73.8
- 10 MiMo-V2-Pro 72.3
内容风格
12 models we track have a score here.
- 1 GPT-5.5 ↑7 92.0
- 2 Kimi K2.6 ↑2 91.0
- 3
ERNIE 5.1 ↓1 89.6
- 4
Qwen 3.6 Max (Preview) ↑1 89.5
- 5
DeepSeek-V4-Pro ↑4 89.0
- 6 Grok 4.20 88.7
- 7
Gemini 3.1 Pro ↓4 88.7
- 8 GLM-5.1 ↓1 88.0
- 9 MiMo-V2-Pro ↑1 87.1
- 10
DeepSeek-V4-Flash ↑1 84.5
内容创意
12 models we track have a score here.
- 1 Kimi K2.6 ↑3 89.4
- 2
ERNIE 5.1 89.3
- 3 Claude Opus 4.7 ↓2 88.7
- 4 Grok 4.20 ↑2 88.2
- 5
DeepSeek-V4-Pro ↑4 87.0
- 6 GPT-5.5 ↑2 85.5
- 7
DeepSeek-V4-Flash ↑4 84.4
- 8 MiMo-V2-Pro ↑2 83.5
- 9
Gemini 3.1 Pro ↓6 83.4
- 10 GLM-5.1 ↓3 83.2
主流题材创作
12 models we track have a score here.
- 1
ERNIE 5.1 ↑1 84.9
- 2 Kimi K2.6 ↑2 82.8
- 3
DeepSeek-V4-Flash ↑8 82.6
- 4
Gemini 3.1 Pro ↓1 82.5
- 5
Qwen 3.6 Max (Preview) 81.7
- 6 Claude Opus 4.7 ↓5 81.1
- 7
DeepSeek-V4-Pro ↑2 80.0
- 8 GLM-5.1 ↓1 78.8
- 9 MiMo-V2-Pro ↑1 77.5
- 10 Grok 4.20 ↓4 75.2
约束创作
12 models we track have a score here.
- 1 GPT-5.5 ↑7 96.2
- 2 Claude Opus 4.7 ↓1 95.4
- 3
Gemini 3.1 Pro 93.9
- 4 Grok 4.20 ↑2 93.0
- 5 GLM-5.1 ↑2 92.5
- 6
Qwen 3.6 Max (Preview) ↓1 92.5
- 7 Kimi K2.6 ↓3 86.0
- 8
ERNIE 5.1 ↓6 84.5
- 9
MiniMax-M2.7 ↑3 76.7
- 10
DeepSeek-V4-Pro ↓1 74.6
事实可靠性
12 models we track have a score here.
- 1 GPT-5.5 ↑7 93.3
- 2 GLM-5.1 ↑5 88.9
- 3
ERNIE 5.1 ↓1 88.9
- 4
Qwen 3.6 Max (Preview) ↑1 86.7
- 5 Claude Opus 4.7 ↓4 84.4
- 6 Kimi K2.6 ↓2 75.6
- 7 Grok 4.20 ↓1 71.1
- 8
Gemini 3.1 Pro ↓5 71.1
- 9 MiMo-V2-Pro ↑1 64.4
- 10
DeepSeek-V4-Pro ↓1 51.1
内容完整性
12 models we track have a score here.
- 1 Grok 4.20 ↑5 100.0
- 2
Gemini 3.1 Pro ↑1 100.0
- 3 MiMo-V2-Pro ↑7 100.0
- 4 GLM-5.1 ↑3 100.0
- 5 Claude Opus 4.7 ↓4 100.0
- 6 Kimi K2.6 ↓2 100.0
- 7
Qwen 3.6 Max (Preview) ↓2 100.0
- 8 GPT-5.5 100.0
- 9
DeepSeek-V4-Flash ↑2 100.0
- 10
DeepSeek-V4-Pro ↓1 100.0
节奏流畅度
12 models we track have a score here.
- 1
DeepSeek-V4-Pro ↑8 92.9
- 2 Kimi K2.6 ↑2 92.0
- 3 GPT-5.5 ↑5 92.0
- 4
ERNIE 5.1 ↓2 91.7
- 5 Grok 4.20 ↑1 90.9
- 6 Claude Opus 4.7 ↓5 89.8
- 7 MiMo-V2-Pro ↑3 87.2
- 8 GLM-5.1 ↓1 86.5
- 9
Gemini 3.1 Pro ↓6 84.6
- 10
Qwen 3.6 Max (Preview) ↓5 84.5
逻辑严谨性
12 models we track have a score here.
- 1 Claude Opus 4.7 88.9
- 2
Gemini 3.1 Pro ↑1 84.4
- 3 Grok 4.20 ↑3 82.2
- 4
ERNIE 5.1 ↓2 77.8
- 5
Qwen 3.6 Max (Preview) 75.6
- 6 GLM-5.1 ↑1 68.9
- 7 Kimi K2.6 ↓3 68.9
- 8
DeepSeek-V4-Pro ↑1 51.1
- 9
DeepSeek-V4-Flash ↑2 42.2
- 10 GPT-5.5 ↓2 40.0
人设一致性
12 models we track have a score here.
- 1 Grok 4.20 ↑5 100.0
- 2
Gemini 3.1 Pro ↑1 100.0
- 3 MiMo-V2-Pro ↑7 100.0
- 4 GLM-5.1 ↑3 100.0
- 5 Kimi K2.6 ↓1 100.0
- 6
Qwen 3.6 Max (Preview) ↓1 100.0
- 7 GPT-5.5 ↑1 100.0
- 8
DeepSeek-V4-Flash ↑3 100.0
- 9
DeepSeek-V4-Pro 96.7
- 10
ERNIE 5.1 ↓8 96.7
场景风格适配性
12 models we track have a score here.
- 1 GPT-5.5 ↑7 88.3
- 2 Grok 4.20 ↑4 86.4
- 3 Kimi K2.6 ↑1 85.4
- 4
Qwen 3.6 Max (Preview) ↑1 84.6
- 5
ERNIE 5.1 ↓3 84.4
- 6
DeepSeek-V4-Pro ↑3 84.4
- 7 Claude Opus 4.7 ↓6 83.9
- 8
Gemini 3.1 Pro ↓5 83.2
- 9 MiMo-V2-Pro ↑1 81.8
- 10 GLM-5.1 ↓3 79.4
语言表现力
12 models we track have a score here.
- 1 Claude Opus 4.7 87.8
- 2 GPT-5.5 ↑6 87.7
- 3
ERNIE 5.1 ↓1 87.6
- 4 Kimi K2.6 87.5
- 5
DeepSeek-V4-Pro ↑4 86.1
- 6 GLM-5.1 ↑1 84.5
- 7
Qwen 3.6 Max (Preview) ↓2 83.7
- 8
Gemini 3.1 Pro ↓5 82.9
- 9 Grok 4.20 ↓3 79.8
- 10 MiMo-V2-Pro 79.5
情绪共鸣
12 models we track have a score here.
- 1 GPT-5.5 ↑7 93.0
- 2 Grok 4.20 ↑4 92.8
- 3 Kimi K2.6 ↑1 92.2
- 4
DeepSeek-V4-Pro ↑5 91.5
- 5 Claude Opus 4.7 ↓4 91.4
- 6
ERNIE 5.1 ↓4 91.3
- 7
Gemini 3.1 Pro ↓4 89.9
- 8 GLM-5.1 ↓1 89.3
- 9 MiMo-V2-Pro ↑1 87.9
- 10
Qwen 3.6 Max (Preview) ↓5 87.7
结构反套路
12 models we track have a score here.
- 1
ERNIE 5.1 ↑1 86.3
- 2 Kimi K2.6 ↑2 84.4
- 3 Claude Opus 4.7 ↓2 83.1
- 4 Grok 4.20 ↑2 82.2
- 5
Gemini 3.1 Pro ↓2 78.8
- 6 GLM-5.1 ↑1 77.4
- 7 MiMo-V2-Pro ↑3 77.2
- 8
Qwen 3.6 Max (Preview) ↓3 76.3
- 9
DeepSeek-V4-Flash ↑2 74.7
- 10
DeepSeek-V4-Pro ↓1 72.3
视角创新
12 models we track have a score here.
- 1
ERNIE 5.1 ↑1 90.3
- 2
DeepSeek-V4-Pro ↑7 88.7
- 3 Kimi K2.6 ↑1 88.5
- 4 GPT-5.5 ↑4 86.8
- 5
DeepSeek-V4-Flash ↑6 86.8
- 6 Claude Opus 4.7 ↓5 85.9
- 7 Grok 4.20 ↓1 84.1
- 8
Qwen 3.6 Max (Preview) ↓3 77.9
- 9 MiMo-V2-Pro ↑1 77.3
- 10 GLM-5.1 ↓3 75.3
认知启发
12 models we track have a score here.
- 1
DeepSeek-V4-Pro ↑8 95.5
- 2 Claude Opus 4.7 ↓1 94.5
- 3 Grok 4.20 ↑3 93.6
- 4 GPT-5.5 ↑4 92.8
- 5 Kimi K2.6 ↓1 92.5
- 6 MiMo-V2-Pro ↑4 91.7
- 7
Gemini 3.1 Pro ↓4 91.0
- 8 GLM-5.1 ↓1 90.8
- 9
ERNIE 5.1 ↓7 89.5
- 10
DeepSeek-V4-Flash ↑1 88.9
IP衍生创作
12 models we track have a score here.
- 1 Kimi K2.6 ↑3 89.6
- 2
DeepSeek-V4-Pro ↑7 88.7
- 3 Grok 4.20 ↑3 87.7
- 4 Claude Opus 4.7 ↓3 87.4
- 5 GPT-5.5 ↑3 86.5
- 6
Gemini 3.1 Pro ↓3 86.2
- 7
ERNIE 5.1 ↓5 86.1
- 8
Qwen 3.6 Max (Preview) ↓3 86.0
- 9
DeepSeek-V4-Flash ↑2 84.2
- 10 MiMo-V2-Pro 84.1
剧本杀
12 models we track have a score here.
- 1
DeepSeek-V4-Flash ↑10 87.5
- 2 Claude Opus 4.7 ↓1 85.8
- 3
Gemini 3.1 Pro 85.5
- 4
ERNIE 5.1 ↓2 85.3
- 5 GLM-5.1 ↑2 83.7
- 6
Qwen 3.6 Max (Preview) ↓1 80.4
- 7
DeepSeek-V4-Pro ↑2 75.7
- 8 Kimi K2.6 ↓4 70.8
- 9 Grok 4.20 ↓3 68.9
- 10 MiMo-V2-Pro 67.6
小说故事
12 models we track have a score here.
- 1
DeepSeek-V4-Pro ↑8 92.2
- 2
DeepSeek-V4-Flash ↑9 86.9
- 3 Kimi K2.6 ↑1 86.9
- 4 MiMo-V2-Pro ↑6 85.5
- 5
ERNIE 5.1 ↓3 84.2
- 6
Gemini 3.1 Pro ↓3 82.7
- 7 Grok 4.20 ↓1 82.1
- 8 GPT-5.5 82.0
- 9 Claude Opus 4.7 ↓8 81.2
- 10
Qwen 3.6 Max (Preview) ↓5 80.8
广告文案
12 models we track have a score here.
- 1 Kimi K2.6 ↑3 86.9
- 2 Claude Opus 4.7 ↓1 86.8
- 3 Grok 4.20 ↑3 86.6
- 4
ERNIE 5.1 ↓2 82.5
- 5 GPT-5.5 ↑3 80.9
- 6
Qwen 3.6 Max (Preview) ↓1 80.6
- 7
Gemini 3.1 Pro ↓4 80.2
- 8 GLM-5.1 ↓1 78.4
- 9
MiniMax-M2.7 ↑3 73.3
- 10 MiMo-V2-Pro 69.8
漫画分镜
12 models we track have a score here.
- 1
DeepSeek-V4-Pro ↑8 87.7
- 2
ERNIE 5.1 85.2
- 3
DeepSeek-V4-Flash ↑8 80.3
- 4
Qwen 3.6 Max (Preview) ↑1 78.9
- 5 MiMo-V2-Pro ↑5 78.6
- 6
Gemini 3.1 Pro ↓3 72.0
- 7 Kimi K2.6 ↓3 70.3
- 8 GLM-5.1 ↓1 68.5
- 9 Claude Opus 4.7 ↓8 58.8
- 10 GPT-5.5 ↓2 57.4
直播脚本
12 models we track have a score here.
- 1
DeepSeek-V4-Pro ↑8 85.1
- 2
Gemini 3.1 Pro ↑1 83.3
- 3 Kimi K2.6 ↑1 83.1
- 4 Grok 4.20 ↑2 81.9
- 5
DeepSeek-V4-Flash ↑6 81.2
- 6
ERNIE 5.1 ↓4 79.9
- 7 Claude Opus 4.7 ↓6 79.3
- 8 MiMo-V2-Pro ↑2 77.3
- 9 GPT-5.5 ↓1 77.3
- 10 GLM-5.1 ↓3 76.5
短视频脚本
12 models we track have a score here.
- 1
ERNIE 5.1 ↑1 85.8
- 2 Kimi K2.6 ↑2 83.4
- 3
DeepSeek-V4-Flash ↑8 82.0
- 4
Gemini 3.1 Pro ↓1 82.0
- 5
Qwen 3.6 Max (Preview) 81.7
- 6 Claude Opus 4.7 ↓5 80.3
- 7 Grok 4.20 ↓1 79.0
- 8 GPT-5.5 78.7
- 9 GLM-5.1 ↓2 76.9
- 10 MiMo-V2-Pro 74.8
网剧剧本
12 models we track have a score here.
- 1 Kimi K2.6 ↑3 91.0
- 2
DeepSeek-V4-Flash ↑9 90.3
- 3 GPT-5.5 ↑5 90.3
- 4
ERNIE 5.1 ↓2 90.1
- 5
Qwen 3.6 Max (Preview) 90.0
- 6 Claude Opus 4.7 ↓5 89.1
- 7
Gemini 3.1 Pro ↓4 88.3
- 8
DeepSeek-V4-Pro ↑1 88.3
- 9 GLM-5.1 ↓2 86.6
- 10 MiMo-V2-Pro 82.2
多约束兼容
12 models we track have a score here.
- 1 Claude Opus 4.7 100.0
- 2 GPT-5.5 ↑6 100.0
- 3
Gemini 3.1 Pro 96.7
- 4 Grok 4.20 ↑2 93.8
- 5 GLM-5.1 ↑2 93.8
- 6
Qwen 3.6 Max (Preview) ↓1 93.3
- 7 Kimi K2.6 ↓3 80.0
- 8
ERNIE 5.1 ↓6 78.7
- 9
MiniMax-M2.7 ↑3 75.9
- 10 MiMo-V2-Pro 59.0
跨题材迁移
12 models we track have a score here.
- 1 GPT-5.5 ↑7 92.3
- 2 Grok 4.20 ↑4 92.2
- 3 Kimi K2.6 ↑1 92.1
- 4
DeepSeek-V4-Pro ↑5 92.0
- 5
Qwen 3.6 Max (Preview) 91.7
- 6 GLM-5.1 ↑1 91.2
- 7
Gemini 3.1 Pro ↓4 91.2
- 8 Claude Opus 4.7 ↓7 90.8
- 9
ERNIE 5.1 ↓7 90.2
- 10 MiMo-V2-Pro 89.0
3 models have held number one on the official leaderboard since 2023
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official SuperCLUE-Writing leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official SuperCLUE-Writing leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Writing and taste
- Design Bench (Moonshot) In use Models make graphic designs from direct prompts or few-shot reference images. An automated grader rates the results as Elo scores. Kimi K3.1 Pro 957
- WritingBench Fading Models complete 1,000 real-world writing requests across 6 domains and 100 subdomains. A judge model scores each reply on five query-specific criteria. GPT-5 83.9
- Creative Writing (MiniMax in-house) Only in system cards Models write creative pieces for real Hailuo AI user requests. Automatic assessments score the writing.
MiniMax-Text-01 81.3%
- DeepSeek Chinese functional writing evaluation Only in system cards Models answer 3,170 everyday Chinese writing prompts such as reports, emails and speeches. Pairwise judging against a baseline model gives a win rate.
DeepSeek-V4-Pro 62.6%
- Slide generation human evaluation (Z.ai) Only in system cards People compare HTML slide decks from two models on content, layout and visual quality. Metric is overall win rate. GLM-5 67.5%
- Task-dependent output homogenization evaluation Only in system cards Models answer prompts several times. A judge scores functional diversity where it is wanted and consistency where one answer is right.
Muse Spark 0.21
Data audit
7 of 9 checks pass, 2 warnings. Look first at the Task access check. We don’t know whether the tasks are public. Every benchmark
Catalog record 1 warning 1 pass
- Warning Task access We don’t know whether the tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
- Pass Independent runs Scores from the official leaderboard (14 entries).
- Pass Models scored 12 models we track, plus 2 other entries on its leaderboard.
- Warning Newest score The newest came 6 months ago, in May 2026, for ERNIE 5.1.
- Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.5 (#21).
- Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on superclueai.com: 12 rows for models we track and 2 other entries at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.