Benchmarks · composite

Arena Text · Excluding ties

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

All results

223 rows

#	Model	Effort	Score
1	Gemini 4 Carbon	high	1550 ±13
2	Gemini 4 Argon	high	1548 ±7
3	Muse Wonder	max	1534 ±9
4	Opus 5.5	high	1525 ±8
5	Fable 5	high	1523 ±5
6	Opus 4.6	high	1522 ±4
7	Fable 5.1	max	1521 ±7
8	Opus 4.7	high	1519 ±5
9	Gemini 3.8 Flash	high	1515 ±6
10	Opus 4.6	max	1515 ±4
11	Muse Spark 1.3	max	1514 ±7
12	Muse Spark 1.2	xhigh	1509 ±8
13	Opus 4.7	max	1508 ±5
14	Muse Spark 1.1	max	1506 ±5
15	Opus 5	high	1504 ±5
16	Gemini 3.7 Flash	high	1504 ±6
17	Opus 5	max	1503 ±6
18	Muse Spark		1502 ±8
19	Gemini 3.1 Pro	high	1501 ±4
20	Gemini 3 Pro	high	1501 ±5
21	Kimi K3	max	1501 ±6
22	GPT-5.6 Sol	xhigh	1500 ±5
23	Qwen 3.8 Max	max	1498 ±6
24	Kimi K3.1	max	1496 ±14
25	GPT-6.1 Sol	max	1496 ±8
26	Gemini 3.6 Flash	high	1495 ±6
27	Opus 4.8	high	1494 ±5
28	GLM 5.3	max	1493 ±6
29	GPT-5.5	high	1492 ±5
30	MiMo V2.6 Pro	max	1489 ±13