Arena Vision

Human preference on prompts with image inputs (multimodal understanding); Bradley-Terry rating.

All results

113 rows

#	Model	Effort	Arena Vision
1	Fable 5	high	1309 ±7
2	Qwen 3.8 Max	max	1300 ±7
3	Opus 4.6	high	1299 ±6
4	Opus 4.7	max	1299 ±7
5	Opus 4.7	high	1298 ±7
6	Gemini 3.7 Flash	high	1295 ±9
7	Muse Spark		1294 ±9
8	Opus 4.6	max	1294 ±6
9	GPT-6.1 Sol	max	1294 ±13
10	Opus 5.5	high	1293 ±12
11	Muse Spark 1.2	xhigh	1293 ±14
12	Gemini 3.8 Flash	high	1291 ±10
13	Muse Spark 1.3	max	1290 ±9
14	Gemini 3 Pro	high	1289 ±7
15	Opus 5	high	1288 ±7
16	Muse Spark 1.3	xhigh	1288 ±10
17	Fable 5.1	max	1288 ±9
18	GPT-5.6 Sol	xhigh	1286 ±7
19	GPT-5.5	xhigh	1286 ±6
20	Opus 4.8	high	1286 ±6
21	GPT-6 Astra	max	1285 ±10
22	GPT-5.4	high	1284 ±6
23	Gemini 3.5 Flash	medium	1282 ±7
24	Gemini 3.5 Flash	high	1282 ±7
25	GPT-5.5	high	1282 ±6
26	Gemini 3.6 Flash	high	1282 ±8
27	Muse Spark 1.1	max	1280 ±7
28	Gemini 3.1 Pro	high	1280 ±5
29	Grok 4.5	max	1279 ±7
30	Opus 4.8	max	1279 ±7