Skip to content

Benchmark

Arena ELO Search

People compare model answers and choose which they prefer. Higher scores are better.

Models measured
44
Scores recorded
45
Sources
1
IQ weight
1.91

Score distribution

1,186.8
Mean
1,195.2
Median
52.8
Std dev
1,133.1
10th percentile
1,246.4
90th percentile
1,006.2
Min
1,006.21,266.3
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.91 for chat models, measured over 42 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Amazon-Web-Search-Agent· LMArena data1,266.3
2Qwen3.8 Max· LMArena data1,260.9
3GPT-5.6 SolxHigh· LMArena data1,256.2
4Claude Opus 4.6Search· LMArena data1,251.6
5GPT 5.5Search· LMArena data1,246.4
6Claude Opus 5High· LMArena data1,238.3
7Claude Fable 5.1High· LMArena data1,236.8
8GPT-6 AstraMax· LMArena data1,230.5
9Claude Opus 4.7· LMArena data1,230
10GPT-6.1 SolMax· LMArena data1,229.6
11Claude Fable 5High· LMArena data1,227.7
12ERNIE 5.1· LMArena data1,227.1
13Claude Sonnet 4.6Search· LMArena data1,220.3
14Grok 4.5· LMArena data1,215.6
15Claude Opus 5.5Medium· LMArena data1,213.6
16Gemini 3.1 Pro PreviewGrounding· LMArena data1,211.4
17Gemini 3 Pro PreviewGrounding· LMArena data1,206.8
18GPT 5.2Search· LMArena data1,206.7
19Grok 4.20 Multi-Agent0309, beta· LMArena data1,205.2
20Claude Opus 4.8· LMArena data1,202.3
21GPT 5.1Search· LMArena data1,199.4
22GPT 5.4Search· LMArena data1,198
23Gemini 3 FlashGrounding· LMArena data1,195.2
24GPT-6.2 AstraMax· LMArena data1,191.4
25Grok 4.20 Beta 1· LMArena data1,190.8
26Claude Sonnet 5Search· LMArena data1,187.3
27Claude Opus 4.5Search· LMArena data1,179.2
28GPT-6.2 SolMax· LMArena data1,179.2
29GPT-6 SolMax· LMArena data1,173.8
30GPT 5.2Search, Non-Reasoning· LMArena data1,171.9
31Grok 4.1 FastSearch· LMArena data1,171.7
32Grok 4 FastSearch· LMArena data1,170.8
33Grok 4.3· LMArena data1,166.1
34Claude Sonnet 4.5Search· LMArena data1,158.8
35GPT-6 LunaMax· LMArena data1,156.8
36Claude Opus 4.1Search· LMArena data1,148.2
37OpenAI o3Search· LMArena data1,144.4
38Gemini 2.5 ProGrounding· LMArena data1,143.3
39Grok 4Search· LMArena data1,141.7
40Sonar Reasoning ProHigh· LMArena data1,138.8
41GPT 5Search· LMArena data1,133.1
42Sonar ProHigh· LMArena data1,129.6
43Claude Opus 4Search· LMArena data1,126.9
44Diffbot-Small-XL· LMArena data1,022
45GPT 4oSearch· LMArena data1,006.2

Data credits: LMArena, CC BY 4.0 (licence)