Lech Mazur Writing: leaderboard
Lech Mazur Writing evaluates models that generate short fiction from constrained creative briefs in a head-to-head format. Text is the input modality. The pinned metric is Comparison Score (points), measuring pairwise preference outcomes from evaluator comparisons of stories written for the same required elements. Source: BenchmarkList’s attributed result table.
Metric: Comparison Score (points). Source: benchmarklist.com. 60 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Fable 5.1 (High) | 3.75 | #10 (Claude Fable 5.1) |
| 2 | Claude Opus 5.5 (High) | 3.69 | #3 (Claude Opus 5.5) |
| 3 | Claude Opus 5 (xHigh) | 3.69 | #16 (Claude Opus 5) |
| 4 | GPT-6 Astra (High) | 3.42 | #6 (GPT-6 Astra) |
| 5 | Claude Opus 5 (High) | 3.32 | #16 (Claude Opus 5) |
| 6 | GLM-5.3 (Max) | 2.95 | #44 (GLM-5.3) |
| 7 | Claude Fable 5 (High) | 2.81 | #20 (Claude Fable 5) |
| 8 | Kimi K3.1 Pro (High) | 2.72 | #17 (Kimi K3.1 Pro) |
| 9 | GPT-5.6 Sol (xHigh) | 2.42 | #22 (GPT-5.6 Sol) |
| 10 | GPT-5.5 (xHigh) | 2.39 | #33 (GPT-5.5) |
| 11 | Kimi K3 | 2.35 | #35 |
| 12 | GPT-5.6 Sol (High) | 2.24 | #22 (GPT-5.6 Sol) |
| 13 | GPT-5.4 (Medium) | 1.92 | #83 (GPT-5.4) |
| 14 | GPT-5.4 (xHigh) | 1.92 | #83 (GPT-5.4) |
| 15 | Claude Opus 4.7 (High) | 1.72 | #51 (Claude Opus 4.7) |
Interactive version: theaggregate.ai/benchmark?slug=lech-mazur-writing · How It Works · Data refreshed daily, snapshot 2026-11-05.