Arena-Hard Creative Writing — leaderboard
250 creative writing prompts auto-evaluated by an ensemble of GPT-4.1 and Gemini 2.5 judges, measuring style, narrative quality, and creative expression.
Metric: Win Rate (%). Source: github.com. Status: saturation imminent. 28 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | O3 (2025-04-16) | 88.8 |
| 2 | Gemini 2.5 Flash | 83.9 |
| 3 | DeepSeek R1 | 77 |
| 4 | Qwen 3 235B A22B | 73.5 |
| 5 | Gemma 3 27B (IT) | 69.9 |
| 6 | GPT-4.1 | 61.5 |
| 7 | QwQ-32B | 60.9 |
| 8 | O1 (2024-12-17) (High) | 59.9 |
| 9 | O4 Mini (2025-04-16) (High) | 58.7 |
| 10 | O1 (2024-12-17) | 56.6 |
| 11 | O4 Mini (2025-04-16) | 55.6 |
| 12 | Qwen 3 32B | 53.3 |
| 13 | GPT-4.5 (Preview) | 51.4 |
| 14 | Gemini 2.0 Flash (001) | 50 |
| 15 | O3 Mini (2025-01-31) (High) | 43 |
Interactive version: theaggregate.ai/benchmark?slug=arena-hard-creative-writing · How the rankings work · Data refreshed daily, snapshot 2026-07-22.