Arena Text — Multi-Turn

Crowdsourced pairwise human preference restricted to multi-turn conversations (style control on); Bradley-Terry rating.

All results

220 rows

#	Model	Effort	Arena Text — Multi-Turn
1	Gemini 4 Argon	high	1553 ±23
2	Opus 4.6	high	1518 ±6
3	Fable 5	high	1516 ±8
4	Opus 4.7	high	1516 ±7
5	Opus 4.7	max	1514 ±7
6	Opus 4.6	max	1512 ±6
7	Muse Spark 1.2	xhigh	1506 ±20
8	Gemini 3.8 Flash	high	1503 ±9
9	Kimi K3	max	1500 ±9
10	Opus 5.5	high	1498 ±21
11	Gemini 3.1 Pro	high	1497 ±5
12	Muse Spark 1.1	xhigh	1497 ±8
13	Opus 4.8	high	1497 ±7
14	Gemini 3 Pro	high	1496 ±8
15	GPT-5.2	xhigh	1495 ±8
16	Opus 4.8	max	1494 ±7
17	GPT-5.4	high	1493 ±7
18	Qwen 3.8 Max	max	1492 ±10
19	Gemini 3.7 Flash	high	1492 ±10
20	GPT-5.6 Sol	xhigh	1488 ±8
21	GPT-5.5	high	1487 ±7
22	Gemini 3.6 Flash	high	1487 ±8
23	Muse Spark 1.3	max	1487 ±13
24	GPT-6.1 Sol	max	1487 ±24
25	Opus 5	max	1486 ±9
26	Opus 4.5	high	1486 ±8
27	Fable 5.1	max	1485 ±15
28	GPT-6 Astra	max	1484 ±16
29	Sonnet 5.5	xhigh	1484 ±22
30	GPT-5.5	xhigh	1484 ±7