Benchmarks · composite

Arena Text · Creative writing

Crowdsourced pairwise human preference on open-ended text prompts (style control on); Bradley-Terry rating. Only the Creative writing category.

All results

215 rows

#	Model	Effort	Score
1	Opus 5.5	high	1520 ±13
2	Gemini 4 Argon	high	1517 ±13
3	Gemini 4 Carbon	high	1515 ±20
4	Fable 5	high	1501 ±7
5	Opus 4.6	high	1499 ±6
6	Gemini 3.7 Flash	high	1490 ±8
7	Opus 4.7	high	1489 ±7
8	Fable 5.1	max	1489 ±11
9	Gemini 3.8 Flash	high	1487 ±7
10	Gemini 3 Pro	high	1485 ±8
11	Opus 4.7	max	1483 ±7
12	Gemini 3.1 Pro	high	1481 ±5
13	Opus 4.6	max	1477 ±6
14	Opus 5	max	1474 ±8
15	Opus 5	high	1472 ±6
16	Gemini 3.6 Flash	high	1471 ±8
17	GPT-5.6 Sol	xhigh	1470 ±7
18	Gemini 3.5 Flash	medium	1470 ±7
19	Opus 4.5	high	1470 ±8
20	Opus 4.8	high	1469 ±7
21	Qwen 3.8 Max	max	1469 ±9
22	Sonnet 5.5	xhigh	1468 ±14
23	Gemini 3.5 Flash	high	1465 ±7
24	Muse Spark 1.3	max	1464 ±10
25	Muse Spark		1464 ±14
26	Grok 4.20	xhigh	1463 ±10
27	Opus 4.8	max	1463 ±6
28	Kimi K3	max	1463 ±8
29	Gemini 3 Flash	high	1457 ±9
30	GLM 5.3	max	1457 ±9