Skip to content

Benchmark

Arena ELO Text to Image - Photorealistic

Higher scores are better.

Models measured
79
Scores recorded
86
Sources
1
IQ weight
1.49

Score distribution

1,156.4
Mean
1,147
Median
113.2
Std dev
1,034.9
10th percentile
1,314.4
90th percentile
922
Min
9221,434.4
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.49 for image-generation models, measured over 68 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 data1,434.4
2GPT-Image-2.5 Flare· LMArena data1,416.5
3GPT-Image-2.6 SunburstMedium· LMArena data1,411.3
4MAI Image 2.7· LMArena data1,388.5
5GPT-Image-2.6 FlareMedium· LMArena data1,387.7
6GPT Image 2Medium· LMArena data1,378.5
7Gemini Nano Banana 2.1· LMArena data1,337.4
8MAI-Image-2.6· LMArena data1,331.2
9Grok Imagine Image 2.0· LMArena data1,314.4
10Reve 2.1· LMArena data1,303.8
11Grok Imagine Image 2.020260801· LMArena data1,291.8
12Reve 2.0· LMArena data1,277
13Muse Image· LMArena data1,274.4
14Seedream 5.0 Pro· LMArena data1,273.3
15Gemini 3 Pro Imagenano-banana-pro, 2K· LMArena data1,259.6
16Gemini 3.1 Flash Imagenano-banana-2, Web Search· LMArena data1,259.6
17Qwen Image 3.0 Pro· LMArena data1,259.4
18Gemini 3.1 Flash-Lite Imagenano-banana-2-lite· LMArena data1,258.8
19Mai Image 2.5· LMArena data1,250.8
20GPT Image 1.5High Fidelity· LMArena data1,246.7
21Gemini 3 Pro Imagenano-banana-pro· LMArena data1,239.8
22Qwen Image 2.1· LMArena data1,231.9
23MiniMax H3-Image· LMArena data1,208
24Qwen Image 2.0 Pro2026-06-22· LMArena data1,206.6
25MAI-Image-2· LMArena data1,202.9
26Ideogram V4Quality· LMArena data1,199.9
27UNI-1.1 Max· LMArena data1,188.2
28UNI-1.1· LMArena data1,182.2
29Recraft V4.1 Utility Pro· LMArena data1,181.7
30Grok Imagine Image· LMArena data1,176
31Grok Imagine Image Pro· LMArena data1,169.1
32Doubao Seedream 4.5· LMArena data1,165.8
33Cosmos3-Super-Text2ImageAgentic· LMArena data1,165.6
34FLUX.2 [max]· LMArena data1,161.1
35Seedream 5.0 Lite· LMArena data1,158.2
36Gemini 2.5 Flash Imagenano-banana· LMArena data1,154.4
37Hunyuan Image V3· LMArena data1,153.4
38Imagen 4 Ultragenerate-001· LMArena data1,152.7
39FLUX.2 [pro]· LMArena data1,151.7
40Doubao Seedream 4-02K· LMArena data1,149.8
41Cosmos3-Super-Text2Image· LMArena data1,148.9
42Reve v1.5· LMArena data1,147.3
43Recraft V4.1 Pro· LMArena data1,147
44FLUX.2 [flex]· LMArena data1,145.5
45FLUX.2 [dev]· LMArena data1,142.9
46Krea v2 Medium· LMArena data1,136.6
47Imagen 4generate-001· LMArena data1,133.9
48Qwen Image2512· LMArena data1,133
49Wan 2.6· LMArena data1,130.6
50Doubao Seedream 4-0High Res, fal· LMArena data1,129.7

Showing 1–50 of 86 results.

Data credits: LMArena, CC BY 4.0 (licence)