Arena Search

Human preference among LLMs with integrated web search; Bradley-Terry rating.

36 models · 36 results · data as of 2026-10-29 · arena.ai ↗

All results

36 rows

#	Model	Effort	Arena Search
1	GPT-5.6 Sol	xhigh	1261 ±6
2	Qwen 3.8 Max	max	1261 ±13
3	Opus 4.6	max	1251 ±5
4	GPT-5.5	xhigh	1247 ±5
5	Fable 5.1	high	1246 ±7
6	Opus 5	high	1242 ±7
7	GPT-6 Astra	max	1234 ±8
8	GPT-6.1 Sol	max	1233 ±9
9	Opus 4.7	max	1230 ±5
10	Fable 5	high	1229 ±7
11	Sonnet 4.6	max	1221 ±4
12	Grok 4.5	max	1216 ±6
13	Opus 5.5	medium	1213 ±7
14	Gemini 3.1 Pro	high	1209 ±5
15	GPT-5.2	xhigh	1207 ±6
16	Gemini 3 Pro	high	1206 ±5
17	Grok 4.20	xhigh	1205 ±5
18	Opus 4.8	max	1204 ±6
19	GPT-5.1	high	1200 ±5
20	GPT-5.4	xhigh	1199 ±5
21	Gemini 3 Flash	high	1196 ±4
22	Sonnet 5	max	1191 ±6
23	Opus 4.5	high	1180 ±6
24	GPT-6 Sol	max	1175 ±7
25	Grok 4.1 Fast	high	1171 ±5
26	Grok 4 Fast	none	1170 ±5
27	Grok 4.3	xhigh	1166 ±5
28	Sonnet 4.5	high	1160 ±5
29	GPT-6 Luna	max	1160 ±7
30	Opus 4.1	high	1148 ±5