Benchmarks · Writing & design
LLM Creative Story-Writing Benchmark
Short stories written to the same detailed creative briefs: 600 to 800 words using 10 required elements, compared in pairs by three judge models from other families. A relative score, where zero is near the middle of the field. From Lech Mazur.
57 models · 62 results · updated 2026-11-12 · github.com ↗
Leaderboard
Fable 5.2 high
More from this benchmark
1 published measuresScore vs any published measure1 measures: sub-scores, splits, counts
47 more models under All models in settings.
All results
62 rows| # | Model | Effort | Score |
|---|---|---|---|
| 1 | Fable 5.2 | high | 3.7 ±0.1 |
| 2 | Fable 5.1 | high | 3.6 ±0.1 |
| 3 | Opus 5 | xhigh | 3.6 ±0.1 |
| 4 | Opus 5.5 | high | 3.6 ±0.1 |
| 5 | GPT-6 Astra | high | 3.3 ±0.1 |
| 6 | Opus 5 | high | 3.2 ±0.1 |
| 7 | GLM 5.3 | max | 2.8 ±0.1 |
| 8 | Fable 5 | high | 2.7 ±0.1 |
| 9 | Kimi K3.1 | high | 2.5 ±0.1 |
| 10 | GPT-5.6 Sol | xhigh | 2.3 ±0.1 |
| 11 | GPT-5.5 | xhigh | 2.3 ±0.1 |
| 12 | Kimi K3 | max | 2.2 ±0.1 |
| 13 | Hy4 | high | 2.2 ±0.1 |
| 14 | GPT-5.6 Sol | high | 2.1 ±0.1 |
| 15 | GPT-5.4 | medium | 1.8 ±0.1 |
| 16 | GPT-5.4 | xhigh | 1.8 ±0.1 |
| 17 | Gemini 4 Carbon | high | 1.8 ±0.1 |
| 18 | Muse Wonder | high | 1.7 ±0.1 |
| 19 | Opus 4.7 | high | 1.6 ±0.1 |
| 20 | Sonnet 4.6 | max | 1.5 ±0.1 |
| 21 | MiMo V2.6 Pro | max | 1.3 ±0.1 |
| 22 | Opus 4.6 | max | 1.0 ±0.2 |
| 23 | Opus 4.8 | xhigh | 0.6 ±0.1 |
| 24 | Muse Spark 1.1 | high | 0.6 ±0.1 |
| 25 | Gemini 4 Argon | high | 0.6 ±0.1 |
| 26 | Muse Spark 1.3 | high | 0.3 ±0.1 |
| 27 | GLM 5.2 | max | 0.3 ±0.1 |
| 28 | DeepSeek V4 Pro | high | 0.3 ±0.1 |
| 29 | Grok 4.7 | high | 0.2 ±0.2 |
| 30 | GPT-5.2 | medium | 0.2 ±0.2 |