Skip to content

Benchmark

Arena ELO Image Edit - Multi Image Edit

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

Models measured
45
Scores recorded
48
Sources
1
IQ weight
1.03

Score distribution

1,269.3
Mean
1,265.8
Median
123.1
Std dev
1,111.2
10th percentile
1,431.5
90th percentile
1,036.1
Min
1,036.11,527.1
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.03 for image-generation models, measured over 37 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-Image-2.5 Sunburst· LMArena data1,527.1
2GPT-Image-2.5 Flare· LMArena data1,475.4
3MAI Image 2.7· LMArena data1,457.9
4GPT Image 2Medium· LMArena data1,455
5GPT-Image-2.6 FlareMedium· LMArena data1,441.6
6Gemini Nano Banana 2.1· LMArena data1,431.5
7Krea 3 Ultra· LMArena data1,430.5
8Seedream 5.0 Pro· LMArena data1,411.6
9Muse Image· LMArena data1,400
10Gemini 3.1 Flash Imagenano-banana-2, Web Search· LMArena data1,372
11MiniMax H3-Image· LMArena data1,366.5
12Gemini 3 Pro Imagenano-banana-pro· LMArena data1,365.7
13Gemini 3 Pro Imagenano-banana-pro, 2K· LMArena data1,364
14ChatGPT Image LatestHigh Fidelity, 20251216· LMArena data1,350.4
15Reve 2.0· LMArena data1,345.9
16GPT Image 1.5High Fidelity· LMArena data1,342.1
17Ideogram V4.5· LMArena data1,335
18UNI-1.1 Max· LMArena data1,307.9
19Doubao Seedream 4.5· LMArena data1,290.1
20Gemini 3.1 Flash-Lite Imagenano-banana-2-lite· LMArena data1,289.9
21Wan 2.7 Image Pro· LMArena data1,289.7
22UNI-1.1· LMArena data1,287.5
23Wan 2.7 Image· LMArena data1,279.6
24Seedream 5.0 Lite· LMArena data1,265.8
25Reve v1.1· LMArena data1,261.7
26Kling Image O1· LMArena data1,253.4
27FLUX.2 [max]· LMArena data1,250.1
28FLUX.2 [pro]· LMArena data1,240.3
29FLUX.2 [flex]· LMArena data1,234.8
30Gemini 2.5 Flash Imagenano-banana· LMArena data1,233.9
31Reve v1· LMArena data1,224
32Wan 2.6 Image· LMArena data1,220.6
33FLUX.2 [klein] 9B· LMArena data1,211.6
34Doubao Seedream 4-0High Res, fal· LMArena data1,209.1
35FLUX.2 [dev]· LMArena data1,202.2
36Doubao Seedream 4-02K· LMArena data1,197.7
37Qwen Image Edit2511· LMArena data1,173.7
38FLUX.2 [klein] 4B· LMArena data1,166.4
39Wan 2.5 Preview· LMArena data1,165.1
40Doubao Seedream 4-0fal· LMArena data1,157.7
41P Image Edit· LMArena data1,145.6
42GPT Image 1· LMArena data1,126.4
43GPT-Image-1 mini· LMArena data1,111.2
44Flux Kontext Pro· LMArena data1,062.9
45Flux Kontext Max· LMArena data1,061.5
46Gemini 2.0 Flash Image Generation· LMArena data1,058
47SeedEdit 3.0· LMArena data1,042
48Flux Kontext Dev· LMArena data1,036.1

Data credits: LMArena, CC BY 4.0 (licence)