Valumigo

LMArena · 图像编辑

定期获取并展示公开的基准测试资料(LMArena 用户投票排名、Epoch AI 测试分数、OpenRouter 使用量排名)。每个榜单均标注数据发布日期和获取日期。这些分数并非本站自行评定。

LMArena 分数来自用户并排查看两个模型的回答,再投票选出更好的一方(Elo 方式)。如果分数差异在置信区间内,可视为水平基本相近。

如何解读此排行榜

衡量什么: 提供图片及更换背景、调整颜色等编辑指令后,通过比较两个模型的结果并投票得出的排名。

如何解读分数: 这是用户更偏好的编辑结果的相对分数,并非分别对指令遵循和原图保留进行评分的指标。

注意事项: 是否适合所需的编辑类型,请使用相同图片和相同指令直接比较确认。

本次排名的解读要点

  • 第 1 名 gpt-image-2.5-sunburst 与第 2 名 gpt-image-2.5-flare 的 95% 置信区间不重叠。在本次统计中,第 1 名领先的证据较为明确。
  • 前 10 个模型中,OpenAI 的模型有 3 个,数量最多。
  • 第 1 名模型获得了 47,289 票,此排行榜包含 50 个模型。

图像编辑 LMArena

发布日期:2026-09-30 · 获取日期:2026-10-04

13201380144015001560
Ogpt-image-2.5-sunburst
1522
Ogpt-image-2.5-flare
1478
Ogpt-image-2 (medium)
1461
xgrok-imagine-image-2.0 (low)
1427
Mmai-image-2.6
1427
Metamuse-image
1403
Mmai-image-2.5
1401
ByteDanceseedream-5.0-pro
1394
xgrok-imagine-image-quality (20260519)
1391
Googlegemini-3-pro-image-2k (nano-banana-pro)
1390
Ochatgpt-image-latest-high-fidelity (20251216)
1389
Googlegemini-3.1-flash-image (nano-banana-2) [web-search]
1387
Googlegemini-3-pro-image-preview (nano-banana-pro)
1386
Rreve-2.1
1374
Ogpt-image-1.5-high-fidelity
1370
Alibabaqwen-image-2.1
1366
Rreve-2.0
1359
Iideogram-4.5
1351
Luni-1.1-max
1334
xgrok-imagine-image
1329

圆点表示得分,横线表示 95% 置信区间。横线重叠意味着水平基本相近。 图表仅显示同一模型各推理设置(high、max 等)中排名最高的一项。表格列出所有设置各自的排名。

各公司最高排名

  • OOpenAIgpt-image-2.5-sunburst#1
  • xxAIgrok-imagine-image-2.0 (low)#4
  • MMicrosoft AImai-image-2.6#5
  • MetaMetamuse-image#6
  • ByteDanceByteDanceseedream-5.0-pro#8
  • GoogleGooglegemini-3-pro-image-2k (nano-banana-pro)#10
  • RRevereve-2.1#14
  • AlibabaAlibabaqwen-image-2.1#16
  • IIdeogramideogram-4.5#18
查看表格(前 50 名)
排名模型开发方分数95% 置信区间投票数
1#1Ogpt-image-2.5-sunburstOpenAI15221517–152847,289
2#2Ogpt-image-2.5-flareOpenAI14781473–148444,792
3#3Ogpt-image-2 (medium)OpenAI14611458–1465293,819
4#4xgrok-imagine-image-2.0 (low)xAI14271422–143236,022
5#5Mmai-image-2.6Microsoft AI14271422–143241,512
6#6Metamuse-imageMeta14031399–1408130,653
7#7Mmai-image-2.5Microsoft AI14011397–1405190,356
8#8ByteDanceseedream-5.0-proByteDance13941391–1397270,020
9#9xgrok-imagine-image-quality (20260519)xAI13911386–139738,256
10#10Googlegemini-3-pro-image-2k (nano-banana-pro)Google13901388–1393638,610
11#11Ochatgpt-image-latest-high-fidelity (20251216)OpenAI13891386–1391617,941
12#12Googlegemini-3.1-flash-image (nano-banana-2) [web-search]Google13871384–1391216,640
13#13Googlegemini-3-pro-image-preview (nano-banana-pro)Google13861383–1389540,778
14#14Rreve-2.1Reve13741368–138020,823
15#15Ogpt-image-1.5-high-fidelityOpenAI13701367–1372642,272
16#16Alibabaqwen-image-2.1Alibaba13661360–137214,617
17#17Rreve-2.0Reve13591352–136519,291
18#18Iideogram-4.5Ideogram13511343–13605,441
19#19Luni-1.1-maxLuma AI13341329–133943,451
20#20xgrok-imagine-imagexAI13291327–1332737,207
21#21Luni-1.1Luma AI13151311–1318177,667
22#22Googlegemini-3.1-flash-lite-image (nano-banana-2-lite)Google13141309–131865,764
23#23Alibabaqwen-image-2.0-pro-2026-06-22Alibaba13031299–130856,781
24#24Thunyuan-image-3.0-instructTencent13031299–1306331,592
25#25Alibabawan2.7-image-proAlibaba13031299–130743,992
26#26ByteDanceseedream-4.5ByteDance13021300–13041,266,531
27#27Alibabawan2.7-imageAlibaba13011297–130544,814
28#28ByteDanceseedream-5.0-liteByteDance12941291–1297446,300
29#29Googlegemini-2.5-flash-image-preview (nano-banana)Google12931291–129511,355,727
30#30ByteDanceseedream-4-2kByteDance12701264–1277210,076
31#31Rreve-v1.1Reve12621259–1264672,675
32#32Bflux-2-maxBlack Forest Labs12621258–1265375,959
33#33Klingkling-image-o1Kling12511248–1255143,692
34#34Bflux-2-proBlack Forest Labs12451243–1248624,352
35#35Alibabaqwen-image-editAlibaba12411239–12442,022,077
36#36Rreve-v1Reve12351230–1239395,314
37#37Alibabaqwen-image-edit-2511Alibaba12341231–1237485,308
38#38Alibabawan2.6-imageAlibaba12311228–1233664,189
39#39Bflux-2-flexBlack Forest Labs12251222–1228443,953
40#40Bflux-2-klein-9bBlack Forest Labs12251222–1228540,833
41#41Bflux-2-devBlack Forest Labs12251221–1229254,450
42#42ByteDanceseedream-4-high-res-falByteDance12171214–12201,229,943
43#43?p-image-edit?12111208–1214359,965
44#44ByteDanceseedream-4-falByteDance12101204–1216152,114
45#45Rreve-v1.1-fastReve12071204–1210553,361
46#46Rreve-edit-fastReve11981194–1202230,289
47#47Bflux-2-klein-4bBlack Forest Labs11871185–1190541,422
48#48Bflux-1-kontext-maxBlack Forest Labs11811178–1183390,877
49#49Alibabawan2.5-i2i-previewAlibaba11811178–1183715,740
50#50Bflux-1-kontext-proBlack Forest Labs11761174–11796,404,790

数据:LMArena leaderboard dataset(CC BY 4.0)。修改:各榜单仅显示前 50 名,分数四舍五入。 https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset公司标志为各公司的商标,仅用于帮助区分(图标:Simple Icons)。数据:Epoch AI — Capabilities & benchmarking(CC BY 4.0)。 https://epoch.ai/benchmarksSource: OpenRouter (openrouter.ai/rankings), as of 2026-10-04. Licensed under CC BY 4.0. https://openrouter.ai/rankings