0

Creative Story-Writing Benchmark

Saturated

Each model writes a short story that must incorporate 10 assigned elements (characters, objects, concepts, attributes, motivations). A panel of LLM judges grades element use, prose, coherence, character and originality. Published by Lech Mazur.

Domain
Writing
Format
Custom
Published
Jan 2025
Official leaderboard
github.com/lechmazur/writing

Cite

Notes

Only stored in your browser.

Attribution

Leaderboard scores
Lech Mazur
Attribution policy →

Top score 96.0% by Claude Opus 5 - 42 models reporting (16 frontier)

Score history

41
0%25%50%75%100%Aug 25Nov 25Feb 26May 26Aug 26gpt-oss-120bGPT-5.2Claude Opus 4.6GPT-5.5Claude Fable 5Claude Opus 5

Top models

42
Creative Story-Writing BenchmarkBar chart with 21 bars. Highest value: Claude Opus 5 at 96.
21 models

Where it's ranked

1

FAQ

What is Creative Story-Writing Benchmark?
Each model writes a short story that must incorporate 10 assigned elements (characters, objects, concepts, attributes, motivations). A panel of LLM judges grades element use, prose, coherence, character and originality. Published by Lech Mazur.
What is the current top score on Creative Story-Writing Benchmark?
The top reported score is 96.0% by Claude Opus 5, across 42 models reporting (16 from frontier labs).