Creative Writing v3: leaderboard
32 creative-writing prompts run 3 times each (96 pieces) from Sam Paech's EQ-Bench (2025); an LLM judge scores a rubric per piece and ranks models by pairwise Elo, tracking slop.
Metric: Elo. Source: eqbench.com. Saturation forecast: Around August 2027. 106 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Fable 5 | 2027.9 | #16 |
| 2 | Claude Opus 4.7 | 2015.8 | #49 |
| 3 | GPT-5.5 | 1861.3 | #31 |
| 4 | Claude Opus 4.8 | 1831.1 | #37 |
| 5 | GPT-5.4 | 1830.3 | #79 |
| 6 | Claude Sonnet 4.6 | 1790.5 | #90 |
| 7 | Claude Opus 4.6 | 1769.9 | #63 |
| 8 | GLM-5.2 | 1643.2 | #73 |
| 9 | GPT-5.2 | 1628.4 | #110 |
| 10 | Kimi K2.6 | 1618.7 | #104 |
| 11 | Claude Sonnet 4.5 | 1607.8 | #142 |
| 12 | GPT-5.3 | 1601.4 | #103 |
| 13 | Claude Opus 4.5 (20251101) | 1598.2 | #74 |
| 14 | O3 | 1584.2 | #124 |
| 15 | Kimi K2 | 1580.1 | #243 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=creative-writing-v3 · How It Works · Data refreshed daily, snapshot 2026-10-20.