Ranking
Best AI models for writing
Writing and long-form ranking of current AI models. Conservative TrueSkill and WritingBench scores from LLM Stats.
| Rank | Model | Score |
|---|---|---|
| 1 | Qwen3-235B-A22B-Thinking-2507 | 88.3 |
| 2 | Qwen3-Next-80B-A3B-Instruct | 87.3 |
| 3 | Qwen3 VL 235B A22B Thinking | 86.7 |
| 4 | Qwen3 VL 32B Thinking | 86.2 |
| 5 | Qwen3 VL 235B A22B Instruct | 85.5 |
| 6 | Qwen3 VL 8B Thinking | 85.5 |
| 7 | Qwen3 VL 30B A3B Thinking | 85.2 |
| 8 | Qwen3-235B-A22B-Instruct-2507 | 85.2 |
| 9 | Qwen3-Next-80B-A3B-Thinking | 84.6 |
| 10 | Qwen3 VL 4B Thinking | 84.0 |
| 11 | Qwen3 VL 8B Instruct | 83.1 |
| 12 | Qwen3 VL 32B Instruct | 82.9 |
| 13 | Qwen3 VL 30B A3B Instruct | 82.6 |
| 14 | Qwen3 VL 4B Instruct | 82.5 |
| 15 | Qwen3 14B | 78.0 |
| 16 |  Kimi K2-Thinking-0905 | 73.8 |
FAQ
Frequently asked questions
What is the best AI model for writing and long-form content in 2026?
This writing board ranks models on LLM Stats conservative TrueSkill with WritingBench as the named eval. Rank 1 is the current writing lead, which may be an open-weight Qwen model even when GPT or Claude lead coding. Use it for drafts and long-form quality, not for LiveCodeBench.
What is WritingBench on an LLM writing ranking?
WritingBench is a writing-quality benchmark LLM Stats uses for this skill board. It scores generated text quality rather than math contests or function calling. zerouter republishes the WritingBench board and does not run the prompts or graders.
Why might Qwen models lead the AI writing leaderboard?
When Alibaba Cloud Qwen rows sit at the top, that is the live WritingBench conservative order from LLM Stats, not a zerouter editorial pick. Open-weight writing leads are common on this table. Confirm the current rank before you assume GPT or Claude is the writing default.
How is the writing ranking different from the research ranking?
WritingBench scores prose quality. MMLU-Pro on the research board scores broad academic multiple-choice skill. A strong writer can sit mid-table on research, so switch boards if you need citation-style knowledge rather than long-form style.
Where does the writing LLM leaderboard data come from?
The writing ranking is republished from the public LLM Stats WritingBench board at llm-stats.com, including conservative TrueSkill where LLM Stats publishes it. zerouter does not run WritingBench or the other evals behind the board. Scores typically refresh about hourly.