Arena Vision · Entity recognition

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

All results

46 rows

#	Model	Effort	Score
1	Gemini 3 Pro	high	1292 ±36
2	Gemini 3 Flash	high	1290 ±30
3	Gemini 3.1 Pro	high	1284 ±29
4	Gemini 3 Flash	minimal	1274 ±30
5	Gemini 3.1 Flash-Lite	high	1271 ±34
6	GPT-5	high	1261 ±33
7	Gemini 2.5 Pro	high	1251 ±22
8	Opus 4.7	high	1242 ±40
9	GPT-5.2	xhigh	1239 ±48
10	Grok 4.20	xhigh	1238 ±41
11	Opus 4.6	max	1237 ±39
12	Grok 4	high	1236 ±36
13	Opus 4.6	high	1236 ±47
14	Kimi K2.5	max	1232 ±33
15	o3	high	1231 ±31
16	GPT-5.1	high	1228 ±49
17	Gemini 2.5 Flash	medium	1221 ±24
18	GPT-5.5	high	1219 ±48
19	GPT-5.5	xhigh	1219 ±52
20	o4 mini	high	1219 ±33
21	Grok 4.1 Fast	high	1217 ±50
22	GPT-5.4	xhigh	1215 ±46
23	Qwen 3.5 397B-A17B	max	1211 ±37
24	GPT-5 Mini	high	1211 ±40
25	Grok 4.3	xhigh	1210 ±45
26	GPT-5.2	high	1201 ±42
27	GPT-5.4 Mini	high	1198 ±41
28	GPT-4.1 Mini	high	1197 ±34
29	Gemma 4 31B	max	1196 ±29
30	GPT-5.4	high	1196 ±46