Skip to content

Benchmark

Arena ELO Document

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

Models measured
40
Scores recorded
46
Sources
1
IQ weight
2.10

Score distribution

1,456.7
Mean
1,460.6
Median
30.4
Std dev
1,417.1
10th percentile
1,496.6
90th percentile
1,402.1
Min
1,402.11,514.8
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.10 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
1Claude Fable 5.1Max· LMArena data1,514.8
2Claude Opus 5High· LMArena data1,508.7
3Claude Opus 4.6High· LMArena data1,507.6
4Claude Opus 4.6· LMArena data1,506.2
5Claude Fable 5High· LMArena data1,496.6
6Claude Opus 4.7High· LMArena data1,495.7
7Claude Opus 4.7· LMArena data1,494.8
8GPT 5.5High· LMArena data1,484.3
9Gemini 3.8 FlashHigh· LMArena data1,481.9
10Claude Sonnet 4.6· LMArena data1,481.9
11GPT 5.5· LMArena data1,480.9
12GPT-5.6 SolxHigh· LMArena data1,480
13Gemini 3.7 FlashHigh· LMArena data1,476.5
14Claude Opus 4.8High· LMArena data1,474.6
15GPT 5.4· LMArena data1,472.9
16Muse Spark 1.3Max· LMArena data1,472.9
17GPT-5.6 TerraxHigh· LMArena data1,467.7
18GPT-6 AstraMax· LMArena data1,467.3
19Claude Opus 4.520251101· LMArena data1,465.7
20Claude Opus 4.8· LMArena data1,465.1
21Muse Spark 1.1· LMArena data1,463.5
22Gemini 3.5 FlashHigh· LMArena data1,461.8
23Gemini 3.5 FlashMedium· LMArena data1,460.6
24Claude Sonnet 5High· LMArena data1,459.5
25Gemini 3.6 FlashHigh· LMArena data1,456.1
26GPT-5.6 LunaxHigh· LMArena data1,453.8
27Grok 4.5· LMArena data1,451.5
28Kimi K2.6· LMArena data1,450.4
29Claude Sonnet 4.520250929· LMArena data1,449.3
30Muse Spark· LMArena data1,443.6
31Gemini 3.1 Pro Preview· LMArena data1,443.3
32Grok 4.6High· LMArena data1,438.1
33Qwen3.7 Plus· LMArena data1,437.4
34MiniMax M3· LMArena data1,436.9
35Gemma 4 31B IT· LMArena data1,434.3
36Gemini 3 Pro Preview· LMArena data1,433.9
37Kimi K2.5Thinking· LMArena data1,430.7
38Gemini 2.5 Pro· LMArena data1,422.8
39Claude Haiku 4.520251001· LMArena data1,420.4
40GLM 5V Turbo· LMArena data1,418.8
41Grok 4.20Reasoning, 0309, beta· LMArena data1,417.1
42Gemini 3 Flash· LMArena data1,414.3
43GPT 5.2High· LMArena data1,403.9
44GPT-5.5 Instant· LMArena data1,403.2
45GPT 5.1· LMArena data1,403
46GPT 5.2· LMArena data1,402.1

Data credits: LMArena, CC BY 4.0 (licence)