WritingBench — leaderboard
Generative writing benchmark covering writing quality, style control, factuality, and instruction following across diverse writing tasks.
Metric: Score (self-reported). Source: benchmarklist.com. Status: saturated. 53 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 | 83.87 |
| 2 | Claude Sonnet 4.5 | 80.71 |
| 3 | O3 | 80.46 |
| 4 | Qwen 3 235B A22B 2507 Instruct | 80.26 |
| 5 | Claude Opus 4.5 | 79.78 |
| 6 | Gemini 2.5 Pro | 79.26 |
| 7 | Nanbeige4-3B-Thinking-2511 | 79.03 |
| 8 | DeepSeek R1 0528 | 78.92 |
| 9 | doubao-seed-1.6-thinking-250615 | 78.76 |
| 10 | Claude Haiku 4.5 | 77.09 |
| 11 | Claude Opus 4.1 | 76.36 |
| 12 | Claude Opus 4 (20250514) (Thinking) | 75.12 |
| 13 | Qwen 3 235B A22B (Thinking) | 74.73 |
| 14 | Grok 4 | 74.65 |
| 15 | Claude Opus 4 | 74.45 |
Interactive version: theaggregate.ai/benchmark?slug=writingbench · How the rankings work · Data refreshed daily, snapshot 2026-07-22.