Longform Writing
EQ-Bench · Coherence and completeness of long stories
What it measures and how
Uses overall_score_100 for plot and expression across eight chapters. Judge preferences are disclosed as a limitation.
How this evidence is used
Ranked: the two writing evaluations share 6% total, without weight increases for question or model counts.
Evaluation results
Fixed rules select one representative configuration per public model. Anonymous test models are omitted. Original ranks are retained, with up to 30 models per evaluation.
| Source rank | Source model name | Raw result | Representative configuration |
|---|---|---|---|
| 1 | claude-opus-5Anthropic | 86.3 | Source default |
| 2 | claude-fable-5-1Anthropic | 85.3 | Source default |
| 3 | claude-opus-5-5Anthropic | 84.6 | Source default |
| 4 | grok-4.7xAI | 83 | Source default |
| 4 | claude-fable-5Anthropic | 83 | Source default |
| 6 | gpt-6-astraOpenAI | 82.8 | Source default |
| 6 | gpt-6-solOpenAI | 82.8 | Source default |
| 6 | muse-spark-1.3Meta | 82.8 | Source default |
| 9 | aion-3.5aion-labs | 82.1 | Source default |
| 10 | claude-opus-4-7Anthropic | 81.8 | Source default |
| 10 | GLM-5.3Z.ai | 81.8 | Source default |
| 12 | gpt-5.6-solOpenAI | 81.7 | Source default |
| 13 | muse-spark-1.2Meta | 81.5 | Source default |
| 14 | claude-opus-4-8Anthropic | 80.8 | Source default |
| 15 | claude-sonnet-4-6Anthropic | 79.9 | Source default |
| 15 | ox-alpha— | 79.9 | Source default |
| 17 | kimi-k3Moonshot AI | 79.6 | Source default |
| 18 | moonshotai/Kimi-K2.6Moonshot AI | 78.5 | Source default |
| 19 | gpt-5.4OpenAI | 78.3 | Source default |
| 19 | claude-sonnet-5Anthropic | 78.3 | Source default |
| 21 | gpt-5.5OpenAI | 78.2 | Source default |
| 22 | gpt-5.6-terraOpenAI | 78 | Source default |
| 23 | zai-org/GLM-5.2Z.ai | 77.9 | Source default |
| 24 | claude-opus-4-6Anthropic | 77.7 | Source default |
| 25 | gemini-3.8-flashGoogle | 76.8 | Source default |
| 26 | deepseek-ai/DeepSeek-V4-ProDeepSeek | 75.6 | Source default |
| 27 | gpt-5.6-lunaOpenAI | 75.5 | Source default |
| 28 | moonshotai/Kimi-K2.5Moonshot AI | 74.9 | Source default |
| 29 | Altworld/Hemmingway-1— | 74.2 | Source default |
| 30 | deepseek-v4.1-flashDeepSeek | 74 | Source default |
Evaluation limits & attribution
Primarily English writing with model judges. Related tasks using the same judge share a budget and do not establish Chinese prose quality.
Data license: MIT · Metadata declaration in the EQ-bench-site README.
Results are published by EQ-Bench. Original scores and the NexGateHub consensus index use different scales and cannot be added directly.