AA-Image-Editing v2.1Artificial Analysis
Ranks image models on editing an image from a text instruction, by Elo from pairwise human preference votes.
New
6 models added in the last 30 days
Nano Banana 2.1, Ideogram 4.5 (High), HunyuanImage 3.0 Instruct, and 3 more
Range | Model | Elo | Samples | Released | API Pricing 1 | |
|---|---|---|---|---|---|---|
| 1 | 1 |  GPT Image 2.5 Sunburst (max) | 1198±1195% interval 1187 to 1209 | 17,275 | Sep 2026 | $210.7 /1k imgs |
| 2 | 2 | GPT Image 2.5 Flare (max) | 1178±1195% interval 1167 to 1189 | 18,125 | Sep 2026 | $210.7 /1k imgs |
| 3 | 3-5 |  MAI-Image-2.6 | 1151±1395% interval 1138 to 1164 | 15,575 | Aug 2026 | $38.9 /1k imgs |
| 4 | 3-5 |  Nano Banana 2.1 | 1150±1495% interval 1136 to 1164 | 6,148 | Oct 2026 | $33.6 /1k imgs |
| 5 | 3-6 | MAI-Image-2.6-Flash | 1140±1495% interval 1126 to 1154 | 9,936 | Sep 2026 | $19.5 /1k imgs |
| 6 | 5-9 | GPT Image 2 (high) | 1131±1495% interval 1117 to 1145 | 13,705 | Apr 2026 | $211.0 /1k imgs |
| 7 | 6-11 |  Muse Image | 1126±1295% interval 1114 to 1138 | 12,273 | Jul 2026 | $10.0 /1k imgs |
| 8 | 6-12 | MAI-Image-2.5 | 1125±1495% interval 1111 to 1139 | 12,706 | Jun 2026 | $48.1 /1k imgs |
| 9 | 6-12 |  Seedream 5.0 Pro | 1124±1495% interval 1110 to 1138 | 12,561 | Jul 2026 | $90.0 /1k imgs |
| 10 | 7-13 | Nano Banana 2 (Gemini 3.1 Flash Image) | 1120±1395% interval 1107 to 1133 | 15,613 | Feb 2026 | $67.0 /1k imgs |
| 11 | 8-14 | MAI-Image-2.5-Pro | 1117±1395% interval 1104 to 1130 | 14,578 | Jul 2026 | $108.5 /1k imgs |
| 12 | 9-14 | GPT Image 1.5 (high) | 1113±1495% interval 1099 to 1127 | 12,665 | Dec 2025 | $133.0 /1k imgs |
| 13 | 9-14 |  Grok Imagine Image 2.0 | 1112±1495% interval 1098 to 1126 | 7,609 | Aug 2026 | $60.0 /1k imgs |
| 14 | 10-14 | Nano Banana Pro (Gemini 3 Pro Image) | 1110±1395% interval 1097 to 1123 | 14,485 | Nov 2025 | $134.0 /1k imgs |
| 15 | 15-18 | MAI-Image-2.5-Flash | 1094±1495% interval 1080 to 1108 | 12,532 | Jun 2026 | $20.0 /1k imgs |
| 16 | 15-19 | grok-imagine-image-quality | 1090±1395% interval 1077 to 1103 | 7,796 | May 2026 | $50.0 /1k imgs |
| 17 | 15-20 |  HunyuanImage 3.5 (Preview) | 1088±1495% interval 1074 to 1102 | 5,595 | Sep 2026 | $24.0 /1k imgs |
| 18 | 15-20 |  Qwen-Image-3.0-Pro | 1087±1495% interval 1073 to 1101 | 12,651 | Jul 2026 | $40.0 /1k imgs |
| 19 | 16-22 | Qwen-Image-2.1  | 1079±1495% interval 1065 to 1093 | 5,597 | Sep 2026 | No API available |
| 20 | 17-23 |  Luma UNI 1 Max | 1077±1495% interval 1063 to 1091 | 5,679 | May 2026 | $100.0 /1k imgs |
| 21 | 19-25 | grok-imagine-image | 1070±1495% interval 1056 to 1084 | 5,285 | Jan 2026 | $20.0 /1k imgs |
| 22 | 19-27 |  Ideogram 4.5 (High) | 1066±1795% interval 1049 to 1083 | 2,514 | Sep 2026 | $220.0 /1k imgs |
| 23 | 20-26 | Qwen-Image-3.0 | 1063±1495% interval 1049 to 1077 | 7,263 | Jul 2026 | $30.0 /1k imgs |
| 24 | 21-27 | HunyuanImage 3.0 Instruct | 1062±1495% interval 1048 to 1076 | 5,269 | Jan 2026 | No API available |
| 25 | 21-28 | Seedream 5.0 Lite | 1060±1495% interval 1046 to 1074 | 5,615 | Feb 2026 | $35.0 /1k imgs |
| 26 | 21-32 | Qwen Image 2.0 Pro (2026-04-22) | 1055±1995% interval 1036 to 1074 | 2,101 | Apr 2026 | $75.0 /1k imgs |
| 27 | 24-31 | Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image) | 1050±1395% interval 1037 to 1063 | 7,068 | Jun 2026 | $33.6 /1k imgs |
| 28 | 25-33 | Wan 2.6 Image | 1046±1495% interval 1032 to 1060 | 5,446 | Dec 2025 | $30.0 /1k imgs |
| 29 | 25-33 |  DreamImage 2.0 | 1046±1495% interval 1032 to 1060 | 5,418 | Jul 2026 | $80.0 /1k imgs |
| 30 | 26-33 | Wan 2.7 Pro | 1045±1495% interval 1031 to 1059 | 5,484 | Apr 2026 | $75.0 /1k imgs |
| 31 | 27-33 | Luma UNI 1 | 1042±1395% interval 1029 to 1055 | 8,697 | May 2026 | $40.4 /1k imgs |
| 32 | 27-34 | Seedream 4.5 | 1040±1495% interval 1026 to 1054 | 4,656 | Dec 2025 | $40.0 /1k imgs |
| 33 | 29-34 | Wan 2.7 | 1037±1495% interval 1023 to 1051 | 6,827 | Apr 2026 | $26.0 /1k imgs |
| 34 | 32-38 | HunyuanImage 3.0 (Tencent Cloud) | 1023±1895% interval 1005 to 1041 | 2,310 | Sep 2025 | $32.0 /1k imgs |
| 35 | 31-40 |  FLUX.2 [dev] Flash | 1021±2195% interval 1000 to 1042 | 1,760 | Dec 2025 | $5.0 /1k imgs |
| 36 | 34-43 | FLUX.2 [dev] Turbo | 1014±2195% interval 993 to 1035 | 1,786 | Dec 2025 | $8.0 /1k imgs |
| 37 | 34-43 |  FLUX.2 [flex] | 1012±2095% interval 992 to 1032 | 1,936 | Nov 2025 | $120.0 /1k imgs |
| 38 | 35-41 | FLUX.2 [klein] 9B | 1011±1495% interval 997 to 1025 | 6,231 | Jan 2026 | $17.0 /1k imgs |
| 39 | 35-43 | FLUX.2 [pro] | 1008±1795% interval 991 to 1025 | 3,246 | Nov 2025 | $45.0 /1k imgs |
| 40 | 36-43 |  Step Image Edit 2 | 1005±1495% interval 991 to 1019 | 4,708 | May 2026 | $3.0 /1k imgs |
| 41 | 38-44 | FLUX.2 [dev] | 1000AnchorFixed at 1000 as the reference point for this board's Elo scale. | 8,061 | Nov 2025 | $24.0 /1k imgs |
| 42 | 37-48 | GPT Image 1 (high) | 995±2095% interval 975 to 1015 | 1,922 | Apr 2025 | $167.0 /1k imgs |
| 43 | 39-46 | FLUX.2 [max] | 995±1495% interval 981 to 1009 | 6,025 | Dec 2025 | $140.0 /1k imgs |
| 44 | 38-49 | Qwen Image Edit Plus 2511 | 990±2195% interval 969 to 1011 | 1,947 | Nov 2025 | $60.0 /1k imgs |
| 45 | 40-50 | Qwen Image Edit Max 2601 | 985±2095% interval 965 to 1005 | 1,990 | Jan 2026 | $75.0 /1k imgs |
| 46 | 43-49 | Nano Banana (Gemini 2.5 Flash Image) | 984±1595% interval 969 to 999 | 4,427 | Aug 2025 | $39.0 /1k imgs |
| 47 | 43-50 |  P-Image-Edit | 982±1595% interval 967 to 997 | 4,393 | Nov 2025 | $10.0 /1k imgs |
| 48 | 41-50 |  Reve V1 (December) | 981±2195% interval 960 to 1002 | 1,994 | Dec 2025 | $40.0 /1k imgs |
| 49 | 44-52 | Qwen Image 2.0 (2026-03-03) | 972±2195% interval 951 to 993 | 2,027 | Mar 2026 | $35.0 /1k imgs |
| 50 | 45-52 | Qwen Image Edit Plus 2509 | 968±2195% interval 947 to 989 | 1,819 | Sep 2025 | $30.0 /1k imgs |
| 51 | 50-54 |  HiDream-O1-Image | 953±1595% interval 938 to 968 | 4,425 | May 2026 | $10.0 /1k imgs |
| 52 | 50-54 |  Kling Image 3.0 Omni | 953±1595% interval 938 to 968 | 4,413 | Feb 2026 | $28.0 /1k imgs |
| 53 | 51-55 | Kling Image O1 | 944±1595% interval 929 to 959 | 4,325 | Dec 2025 | $28.0 /1k imgs |
| 54 | 50-56 |  Vidu Q2 | 944±2295% interval 922 to 966 | 1,865 | Nov 2025 | $40.0 /1k imgs |
| 55 | 53-58 | FLUX.2 [klein] 4B | 933±1995% interval 914 to 952 | 2,831 | Jan 2026 | $15.0 /1k imgs |
| 56 | 53-60 | GPT Image 1 Mini (medium) | 925±2395% interval 902 to 948 | 1,676 | Oct 2025 | $11.0 /1k imgs |
| 57 | 54-61 | FLUX.1 Kontext [max] | 919±2395% interval 896 to 942 | 1,802 | May 2025 | $80.0 /1k imgs |
| 58 | 55-62 |  FIBO Edit | 913±2395% interval 890 to 936 | 1,792 | Jan 2026 | $40.0 /1k imgs |
| 59 | 55-62 | FLUX.1 Kontext [pro] | 911±2395% interval 888 to 934 | 1,730 | May 2025 | $40.0 /1k imgs |
| 60 | 56-63 | SeedEdit 3.0 | 906±2395% interval 883 to 929 | 1,727 | Jun 2025 | $0.0 /1k imgs |
| 61 | 58-64 | FLUX.1 Kontext [dev] | 893±2495% interval 869 to 917 | 1,635 | Jun 2025 | $25.0 /1k imgs |
| 62 | 58-64 | Qwen Image Edit | 892±2395% interval 869 to 915 | 1,824 | Aug 2025 | $30.0 /1k imgs |
| 63 | 59-64 |  LongCat Image | 888±2495% interval 864 to 912 | 1,702 | Dec 2025 | $130.0 /1k imgs |
| 64 | 62-64 | HiDream-O1-Image-Dev | 879±1695% interval 863 to 895 | 4,378 | May 2026 | $5.0 /1k imgs |
| 65 | 65-67 |  GLM-Image | 816±2795% interval 789 to 843 | 1,749 | Jan 2026 | $50.0 /1k imgs |
| 66 | 65-68 |  OmniGen V2 | 800±2695% interval 774 to 826 | 1,842 | Jun 2025 | $150.0 /1k imgs |
| 67 | 65-68 | Bagel | 789±3095% interval 759 to 819 | 1,592 | May 2025 | $100.0 /1k imgs |
| 68 | 66-68 | HiDream-E1.1 | 777±3195% interval 746 to 808 | 1,514 | Jul 2025 | $60.0 /1k imgs |
| 69 | 69 | Step1X-Edit | 609±4495% interval 565 to 653 | 1,694 | Apr 2025 | $30.0 /1k imgs |
1 API Pricing reflects the cost to generate 1,000 images on the model creator's API at 1024x1024 with the model's default settings
Frequently Asked Questions
The top models on AA-Image-Editing v2.1 by Elo rating are: 1. GPT Image 2.5 Sunburst (max) (Elo 1198), 2. GPT Image 2.5 Flare (max) (Elo 1178), 3. MAI-Image-2.6 (Elo 1151), 4. Nano Banana 2.1 (Elo 1150), and 5. MAI-Image-2.6-Flash (Elo 1140). Rankings are based on blind votes from our validated private evaluators.
Models are ranked using an Elo rating system derived from blind comparisons judged by our validated private evaluators. Voters compare edited outputs from the same input image and editing instruction, choosing the result they prefer. Higher Elo scores indicate a model is preferred more often.