Skip to content

Benchmark

Arena ELO Search Factuality

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

Models measured
42
Scores recorded
43
Sources
1
IQ weight
2.00

Score distribution

1,192.5
Mean
1,198.1
Median
33.2
Std dev
1,143.9
10th percentile
1,238.4
90th percentile
1,129.1
Min
1,129.11,248.4
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.00 for chat models, measured over 40 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1GPT-5.6 SolxHigh· LMArena data1,248.4
2Amazon-Web-Search-Agent· LMArena data1,246.7
3GPT 5.5Search· LMArena data1,246.6
4Qwen3.8 Max· LMArena data1,241.5
5GPT-6 AstraMax· LMArena data1,238.4
6GPT-6.1 SolMax· LMArena data1,236.2
7Claude Opus 4.6Search· LMArena data1,227.3
8GPT 5.2Search· LMArena data1,223.8
9Grok 4.5· LMArena data1,220.1
10ERNIE 5.1· LMArena data1,214.3
11Claude Opus 5High· LMArena data1,213.9
12Claude Fable 5.1High· LMArena data1,213.7
13GPT 5.4Search· LMArena data1,212.1
14Claude Fable 5High· LMArena data1,211.6
15Claude Opus 4.7· LMArena data1,210.6
16Claude Opus 5.5Medium· LMArena data1,206.5
17GPT 5.1Search· LMArena data1,203.5
18Claude Sonnet 4.6Search· LMArena data1,201.8
19GPT 5.2Search, Non-Reasoning· LMArena data1,198.8
20GPT-6 SolMax· LMArena data1,198.4
21Gemini 3.1 Pro PreviewGrounding· LMArena data1,198.3
22Grok 4.20 Multi-Agent0309, beta· LMArena data1,198.1
23Grok 4.20 Beta 1· LMArena data1,195.9
24Claude Opus 4.8· LMArena data1,195.1
25Gemini 3 Pro PreviewGrounding· LMArena data1,193.5
26GPT-6.2 AstraMax· LMArena data1,192.5
27GPT-6 LunaMax· LMArena data1,183.4
28GPT-6.2 SolMax· LMArena data1,183.1
29Claude Opus 4.5Search· LMArena data1,180.4
30Gemini 3 FlashGrounding· LMArena data1,179.1
31Claude Sonnet 5Search· LMArena data1,178.6
32Grok 4.3· LMArena data1,175.1
33Gemini 2.5 ProGrounding· LMArena data1,163
34Claude Sonnet 4.5Search· LMArena data1,161.4
35Sonar Reasoning ProHigh· LMArena data1,155.4
36Claude Opus 4.1Search· LMArena data1,154.4
37Grok 4.1 FastSearch· LMArena data1,152.4
38Grok 4 FastSearch· LMArena data1,146.7
39GPT 5Search· LMArena data1,143.9
40Claude Opus 4Search· LMArena data1,139.2
41Sonar ProHigh· LMArena data1,132.3
42Grok 4Search· LMArena data1,130.4
43OpenAI o3Search· LMArena data1,129.1

Data credits: LMArena, CC BY 4.0 (licence)