Arena Vision · OCR

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

All results

98 rows

#	Model	Effort	Arena Vision · OCR
1	Fable 5	high	1326 ±7
2	Gemini 3.8 Flash	high	1316 ±11
3	Qwen 3.8 Max	max	1316 ±8
4	Opus 4.7	max	1314 ±7
5	Opus 4.6	high	1314 ±7
6	Opus 4.6	max	1312 ±6
7	Opus 4.7	high	1311 ±7
8	Gemini 3.7 Flash	high	1308 ±10
9	Muse Spark 1.3	max	1307 ±10
10	Muse Spark 1.2	xhigh	1307 ±17
11	Opus 5.5	high	1306 ±14
12	Opus 5	high	1304 ±7
13	Fable 5.1	max	1303 ±10
14	Muse Spark		1302 ±10
15	Gemini 3 Pro	high	1302 ±8
16	Gemini 3.6 Flash	high	1301 ±8
17	Muse Spark 1.3	xhigh	1300 ±11
18	GPT-6 Astra	max	1299 ±12
19	Gemini 3.5 Flash	medium	1299 ±7
20	GPT-5.5	xhigh	1299 ±6
21	Opus 4.8	high	1299 ±7
22	GPT-5.4	high	1299 ±6
23	GPT-5.6 Sol	xhigh	1297 ±8
24	GPT-5.5	high	1297 ±6
25	GPT-6.1 Sol	max	1297 ±15
26	Muse Spark 1.1	max	1295 ±7
27	Gemini 3.1 Pro	high	1294 ±5
28	Opus 4.8	max	1293 ±7
29	Gemini 3.5 Flash	high	1292 ±7
30	Sonnet 4.6	max	1291 ±6