Best AI for Creative Writing: Arena Rankings
claude-fable-5.1-max leads Arena creative writing, with 9 more models inside its interval. Some models that rank high overall fall a long way here.
The signal
- LaunchesClaude Fable 5.1 (1 Sept 2026): claude-fable-5.1-max is #1 in creative writing on 1,233 votes.
- LaunchesMuse Spark 1.3 (2 Sept 2026): muse-spark-1.3-max is #10 overall and #25 in creative writing.
- ArenaOpenAI's best creative writing model ranks #22; its best overall model ranks #26.
Anthropic has 7 of the top 10 creative writing models and Google 3. Meta's muse-spark-1.2 (xHigh) is #11 overall but #31 in creative writing.
Top 15 models in the Arena creative writing category, 13 Sept 2026. Bar = Arena points above the #50 creative writing model (qwen3.6-max-preview, 1435). Overall rank in the table and tooltips.
Arena creative writing is a category of the Arena text board: blind human votes on prompts that ask for stories, poems, scripts and similar work. It measures reader preference, not originality or sales. Ranks in the table and tooltips are recomputed over the 400 models present on both boards, so the two ranks compare like with like. 10 models have a 95% interval that reaches the leader's lower bound (1490). Latest snapshots on or before 16 Sept 2026.
Show the data (15 rows)
| Writing rank | Model | Lab | Rating | 95% CI | Overall rank |
|---|---|---|---|---|---|
| 1 | claude-fable-5.1-max | Anthropic | 1508 | 1490–1526 | 1 |
| 2 | claude-opus-4-6-high | Anthropic | 1505 | 1498–1511 | 4 |
| 3 | gemini-3.8-flash-high | 1498 | 1479–1517 | 6 | |
| 4 | claude-fable-5 | Anthropic | 1496 | 1488–1505 | 7 |
| 5 | gemini-3.7-flash-high | 1492 | 1473–1510 | 8 | |
| 6 | claude-opus-5-max | Anthropic | 1492 | 1482–1501 | 2 |
| 7 | claude-opus-5-high | Anthropic | 1488 | 1481–1496 | 3 |
| 8 | claude-opus-4-7-high | Anthropic | 1485 | 1478–1492 | 9 |
| 9 | claude-opus-4-6 | Anthropic | 1484 | 1478–1491 | 5 |
| 10 | gemini-3.1-pro-preview | 1482 | 1476–1487 | 16 | |
| 11 | qwen3.8-max | Alibaba | 1481 | 1470–1492 | 14 |
| 12 | gemini-3-pro | 1481 | 1472–1489 | 17 | |
| 13 | claude-opus-4-7 | Anthropic | 1478 | 1471–1485 | 12 |
| 14 | gemini-3.5-flash-high | 1469 | 1461–1477 | 13 | |
| 15 | gemini-3.5-flash-medium | 1468 | 1459–1476 | 19 |
The Mode read
Anthropic released Claude Fable 5.1 on 1 Sept 2026, and Meta released Muse Spark 1.3 the next day. On the overall board they now rank #1 and #10. Creative writing is a separate test, so which AI is best for creative writing?
claude-fable-5.1-max leads, but 9 other models are inside its interval, so the top is a group, not a single winner. Anthropic holds 7 of the top 10 and Google 3. The overall board does not predict this list well for every lab. Meta has 3 models in the overall top 15 that rank at least 10 places lower in creative writing; muse-spark-1.2 (xHigh) drops from #11 to #31.
Writing is where a model's voice is part of the product. Readers notice it, and they prefer some voices to others. But a voice can be copied by the next release, and the 10 leaders here are within 27 points of each other. What a writing tool can keep is the writer: the drafts, the style guide, and the readers who come back for the next chapter.
A model can copy a voice. It cannot copy a writer's readers.
What we would back: Writing tools that sit where writers already publish and get paid, not another blank text box.