跳到正文

模型能力评测

重点更新

图像编辑 · Arena ↗
59 个模型
图像编辑公开偏好成绩,2026-10-06
模型/机构Arena 评分对战票数
uni-1.1-max#21 · luma-ai
1334.21329–133943,446
1329.01327–1331756,681
uni-1.1#23 · luma-ai
1315.21312–1319188,834
1313.51309–131865,669
1303.21299–130856,723
1302.71299–1306331,423
wan2.7-image-pro#27 · alibaba
1302.61299–130743,974
seedream-4.5#28 · bytedance
1301.91300–13041,294,217
wan2.7-image#29 · alibaba
1301.21297–130544,796
seedream-5.0-lite#30 · bytedance
1294.11291–1297459,744
1293.21291–129511,383,435
seedream-4-2k#32 · bytedance
1270.21264–1277210,059
reve-v1.1#33 · reve
1261.81259–1265672,625
flux-2-max#34 · bfl
1261.71259–1265375,861
kling-image-o1#35 · kling
1251.21248–1255143,639
flux-2-pro#36 · bfl
1245.41243–1248637,986
qwen-image-edit#37 · alibaba
1241.51239–12442,021,952
reve-v1#38 · reve
1234.51230–1239395,313
qwen-image-edit-2511#39 · alibaba
1234.01231–1237485,078
wan2.6-image#40 · alibaba
1230.61228–1233677,385
数据来源

Arena · CC BY 4.0 · 固定版本

用户偏好评分,非正确率。下方小字为置信区间;临时成绩状态见原榜。本站选取 overall 分类、翻译字段并舍入显示,原始值保留。

每日 08:30 同步 · 最近成功:2026/10/11 08:26(北京时间)

历史专项测试:MMLU-Pro / GenEval