图像编辑能力排行榜
本榜收录图像编辑(image-edit)维度在榜的 63 个模型,按 Elo 得分降序排列,当前榜首为 GPT-Image-2.5 Sunburst(OpenAI),得分 1517.4。该维度考察模型在给定原图与编辑指令下改造图片的能力——局部重绘、物体替换、风格迁移、背景处理等,适合设计、电商与内容运营场景参考。
图像编辑 · 完整榜单(63 个模型)
按 Elo 得分降序| 筛选排名 | 排名 | 模型 | 厂商 / 国家 | Elo 得分 | 置信区间 | 投票数 |
|---|---|---|---|---|---|---|
| 1 | GPT-Image-2.5 Sunburst |
OpenAI
|
1517.4 | ±4.5 | 125,387 | |
| 2 | GPT-Image-2.5 Flare |
OpenAI
|
1483.1 | ±4.4 | 122,435 | |
| 3 | GPT-Image-2.6 Sunburst (Medium) |
OpenAI
|
1475.8 | ±10.1 | 3,843 | |
| 4 | MAI Image 2.7 Preview |
Microsoft AI
|
1467.2 | ±6.1 | 19,682 | |
| 5 | GPT-Image-2.6 Flare (Medium) |
OpenAI
|
1462.0 | ±9.9 | 3,852 | |
| 6 | gpt-image-2 (medium) |
OpenAI
|
1459.9 | ±2.9 | 368,928 | |
| 7 | Grok Imagine Image 2.0 |
SpaceXAI
|
1438.4 | ±5.5 | 37,184 | |
| 8 | Gemini Nano Banana 2.1 |
Google
|
1427.2 | ±5.8 | 12,864 | |
| 9 | MAI Image 2.6 |
Microsoft AI
|
1426.5 | ±4.5 | 89,086 | |
| 10 | grok-imagine-image-2.0 (20260801) |
SpaceXAI
|
1416.5 | ±4.1 | 127,412 | |
| 11 | Krea 3 Ultra |
Krea
|
1411.1 | ±10.9 | 2,924 | |
| 12 | MAI Image 2.5 |
Microsoft AI
|
1400.4 | ±3.7 | 190,033 | |
| 13 | Muse Image |
Meta
|
1399.3 | ±3.7 | 204,472 | |
| 14 | Seedream 5.0 Pro |
Bytedance
|
1391.4 | ±3.1 | 353,731 | |
| 15 | grok-imagine-image-quality (20260519) |
SpaceXAI
|
1390.7 | ±5.6 | 38,200 | |
| 16 | Gemini 3 Pro Image (nano-banana-pro, 2K) |
Google
|
1390.2 | ±2.5 | 706,596 | |
| 17 | gemini-3.1-flash-image (nano-banana-2) [web-search] |
Google
|
1386.6 | ±3 | 283,171 | |
| 18 | chatgpt-image-latest-high-fidelity (20251216) |
OpenAI
|
1386.4 | ±2.6 | 678,721 | |
| 19 | gemini-3-pro-image-preview (nano-banana-pro) |
Google
|
1385.2 | ±2.9 | 540,474 | |
| 20 | Reve 2.1 |
Reve
|
1373.5 | ±6.1 | 20,800 | |
| 21 | GPT-Image-1.5 (High Fidelity) |
OpenAI
|
1367.9 | ±2.5 | 704,630 | |
| 22 | Qwen Image 2.1 |
Alibaba
|
1363.9 | ±5.9 | 41,522 | |
| 23 | Reve 2.0 |
Reve
|
1358.2 | ±6.5 | 19,291 | |
| 24 | Ideogram 4.5 |
Ideogram
|
1345.3 | ±4.2 | 100,744 | |
| 25 | UNI-1.1 Max |
Luma AI
|
1333.9 | ±4.9 | 43,463 | |
| 26 | Grok Imagine Image |
SpaceXAI
|
1328.5 | ±2.4 | 860,575 | |
| 27 | UNI 1.1 |
Luma AI
|
1316.1 | ±3.1 | 248,603 | |
| 28 | gemini-3.1-flash-lite-image (nano-banana-2-lite) |
Google
|
1313.0 | ±4.7 | 65,601 | |
| 29 | Seedream 4.5 |
Bytedance
|
1302.6 | ±2.1 | 1,406,275 | |
| 30 | Qwen-Image-2.0-Pro (2026-06-22) |
Alibaba
|
1302.5 | ±4.6 | 56,679 | |
| 31 | Wan 2.7 Image Pro |
Alibaba
|
1302.1 | ±4.1 | 43,965 | |
| 32 | HunyuanImage 3.0 Instruct |
Tencent
|
1301.9 | ±3.3 | 331,407 | |
| 33 | Wan 2.7 Image |
Alibaba
|
1300.7 | ±3.9 | 44,779 | |
| 34 | Seedream 5.0 Lite |
Bytedance
|
1294.1 | ±2.5 | 532,507 | |
| 35 | gemini-2.5-flash-image-preview (nano-banana) |
Google
|
1291.5 | ±2 | 11,490,089 | |
| 36 | Seedream 4.0 (2K) |
Bytedance
|
1266.7 | ±6.9 | 210,028 | |
| 37 | Reve v1.1 |
Reve
|
1261.5 | ±2.7 | 672,512 | |
| 38 | FLUX.2 [max] |
Black Forest Labs
|
1261.5 | ±3.1 | 375,769 | |
| 39 | Kling Image O1 |
KlingAI
|
1250.5 | ±3.7 | 143,662 | |
| 40 | FLUX.2 [pro] |
Black Forest Labs
|
1248.1 | ±2.5 | 694,887 | |
| 41 | Qwen Image Edit |
Alibaba
|
1241.0 | ±2.6 | 2,021,513 | |
| 42 | Reve v1 |
Reve
|
1234.5 | ±5.0 | 395,026 | |
| 43 | Qwen-Image-Edit (2511) |
Alibaba
|
1233.1 | ±2.8 | 484,888 | |
| 44 | Wan 2.6 Image |
Alibaba
|
1231.2 | ±2.5 | 734,821 | |
| 45 | FLUX.2 [flex] |
Black Forest Labs
|
1224.5 | ±3.1 | 443,848 | |
| 46 | FLUX.2 [klein] 9B |
Black Forest Labs
|
1224.3 | ±2.9 | 540,656 | |
| 47 | FLUX.2 [dev] |
Black Forest Labs
|
1224.1 | ±4.0 | 254,036 | |
| 48 | Seedream 4.0 (High Res, fal) |
Bytedance
|
1215.9 | ±2.8 | 1,229,454 | |
| 49 | P Image Edit |
|
1209.9 | ±3.1 | 359,799 | |
| 50 | Seedream 4.0 (fal) |
Bytedance
|
1209.7 | ±6.1 | 152,104 | |
| 51 | Reve v1.1 Fast |
Reve
|
1206.6 | ±3.0 | 553,188 | |
| 52 | Reve Edit Fast |
Reve
|
1198.0 | ±4.1 | 230,047 | |
| 53 | FLUX.2 [klein] 4B |
Black Forest Labs
|
1187.1 | ±2.9 | 541,169 | |
| 54 | FLUX.1 Kontext [max] |
Black Forest Labs
|
1181.9 | ±2.8 | 390,856 | |
| 55 | Wan 2.5 Preview |
Alibaba
|
1180.3 | ±2.6 | 723,629 | |
| 56 | FLUX.1 Kontext [pro] |
Black Forest Labs
|
1177.0 | ±2.7 | 6,404,175 | |
| 57 | FLUX.1 Kontext [dev] |
Black Forest Labs
|
1150.1 | ±3.5 | 3,639,566 | |
| 58 | SeedEdit 3.0 |
Bytedance
|
1138.9 | ±3.8 | 4,918,155 | |
| 59 | GPT Image 1 |
OpenAI
|
1138.6 | ±2.7 | 2,869,445 | |
| 60 | GPT-Image-1 mini |
OpenAI
|
1123.8 | ±2.9 | 682,613 | |
| 61 | Gemini 2.0 Flash Preview Image Generation |
Google
|
1081.6 | ±2.4 | 4,942,634 | |
| 62 | BAGEL |
Bytedance
|
1026.4 | ±5.8 | 13,488 | |
| 63 | Step1X Edit |
StepFun
|
996.0 | ±4.6 | 155,418 |
11 个能力维度速览
每个维度独立成榜、独立收录,下为该维度当前榜首与在榜模型数,点击可直达。
本榜国家与厂商分布
图像编辑维度:63 个模型,来自 15 家厂商(按数量列前 8 家)
本维度在榜 63 个模型,来自 15 家厂商,厂商分布较分散,数量最多的 OpenAI 也只占 14%。国家分布上,中国厂商 19 席、美国厂商 24 席;中国厂商占本维度 30%,与全部 11 个维度的整体水平(32%)大致相当。榜首 GPT-Image-2.5 Sunburst(OpenAI)得分 1517.4,领先第 2 名 34 分;其中 2 个模型置信区间在 ±11 以上(对战样本偏少),名次仍有变动空间。
榜单说明
排序依据:按 Elo 得分降序。得分由真实用户在不知模型身份的前提下对两组回答二选一(盲评对战),经 Elo 算法累积计算, 代表相对胜率而非绝对能力值。置信区间越小表示对战样本越充足;投票数为该模型的累计有效对战场次。 不同维度之间的分数不可横向比较(例如编程维度榜首约 1800 分、文本对话维度榜首约 1506 分,并不代表编程模型更强)。 国家标识依据模型归属厂商判定,同一厂商的多个版本归为同一国家。 数据来源:arena-ai-leaderboards 公开接口,最近抓取时间 2026-11-08 00:25:17,本站每日抓取快照并留存历史。 榜单仅反映真实用户偏好,不构成技术评测结论。 相关榜单:社区榜单(工具热度榜与开源模型下载榜)、 模型库(按厂商 / 参数 / 下载量筛选)。
O
M
S
G
M
A
T
K