Best AI for image generation
Human preference across generation and editing, rated blind, pairwise.
Aldena runs these models inside your team's projects. See what each one costs.
| rank | model | vendor | composite | benchmarks | elo | elo | elo |
|---|---|---|---|---|---|---|---|
| 1 | GPT Image 2.5 Sunburstopenai/gpt-image-2.5-sunburst | OpenAI | 81.9 | 3 of 3 benchmarks | 1428 | 1520 | 1525 |
| 2 | GPT Image 2.5 Flareopenai/gpt-image-2.5-flare | OpenAI | 80.6 | 3 of 3 benchmarks | 1400 | 1485 | 1476 |
| 3 | GPT Image 2 (medium)openai/gpt-image-2:medium | OpenAI | 78.9 | 3 of 3 benchmarks | 1382 | 1461 | 1455 |
| 4 | GPT Image 2.6 Flare (medium)openai/gpt-image-2.6-flare:medium | OpenAI | 78.4 | 3 of 3 benchmarks | 1383 | 1460 | 1439 |
| 5 | GPT Image 2.6 Sunburst (medium)openai/gpt-image-2.6-sunburst:medium | OpenAI | 76.2 | 2 of 3 benchmarks | 1406 | 1477 | |
| 6 | Nano Banana 2.1google/gemini-nano-banana-2.1 | 75.7 | 3 of 3 benchmarks | 1328 | 1427 | 1431 | |
| 7 | MAI-Image-2.6microsoft/mai-image-2.6 | Microsoft AI | 73.2 | 2 of 3 benchmarks | 1330 | 1428 | |
| 8 | Grok Imagine Image 2.0x-ai/grok-imagine-image-2.0 | SpaceXAI | 73.0 | 2 of 3 benchmarks | 1319 | 1437 | |
| 9 | Muse Imagemeta/muse-image | Meta | 72.7 | 3 of 3 benchmarks | 1266 | 1400 | 1399 |
| 10 | Krea 3 Ultrakrea/krea-3-ultra | Krea | 71.6 | 2 of 3 benchmarks | 1411 | 1433 | |
| 11 | Seedream 5.0 Probytedance-seed/seedream-5-0-pro | ByteDance Seed | 71.6 | 3 of 3 benchmarks | 1249 | 1392 | 1412 |
| 12 | Nano Banana 2 (Gemini 3.1 Flash Image)google/gemini-3.1-flash-image | 70.3 | 3 of 3 benchmarks | 1257 | 1387 | 1370 | |
| 13 | MAI-Image-2.5microsoft/mai-image-2.5 | Microsoft AI | 69.3 | 2 of 3 benchmarks | 1253 | 1400 | |
| 14 | Nano Banana Pro (Gemini 3 Pro Image) (2K)google/gemini-3-pro-image:2k | 68.8 | 3 of 3 benchmarks | 1246 | 1391 | 1364 | |
| 15 | Nano Banana Pro (Gemini 3 Pro Image Preview)google/gemini-3-pro-image-preview | 67.7 | 3 of 3 benchmarks | 1231 | 1385 | 1366 | |
| 16 | Reve 2.0reve/reve-2.0 | Reve | 67.6 | 3 of 3 benchmarks | 1268 | 1359 | 1345 |
| 17 | Reve 2.1reve/reve-2.1 | Reve | 67.5 | 2 of 3 benchmarks | 1302 | 1373 | |
| 18 | GPT Image 1.5 High Fidelityopenai/gpt-image-1.5-high-fidelity | OpenAI | 65.5 | 3 of 3 benchmarks | 1236 | 1367 | 1342 |
| 19 | Chatgpt Image High Fidelityopenai/chatgpt-image-high-fidelity | OpenAI | 65.1 | 2 of 3 benchmarks | 1387 | 1351 | |
| 20 | Qwen Image 3.0 Proqwen/qwen-image-3.0-pro | Qwen | 64.7 | 1 of 3 benchmarks | 1254 | ||
| 21 | Grok Imagine Image Qualityx-ai/grok-imagine-image-quality | SpaceXAI | 62.9 | 1 of 3 benchmarks | 1391 | ||
| 22 | Uni-1.1 (max)luma/uni-1.1:max | Luma AI | 62.5 | 3 of 3 benchmarks | 1188 | 1333 | 1308 |
| 23 | Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image)google/gemini-3.1-flash-lite-image | 61.9 | 3 of 3 benchmarks | 1249 | 1313 | 1289 | |
| 24 | Ideogram 4.0 Qualityideogram/ideogram-4.0-quality | Ideogram | 61.8 | 1 of 3 benchmarks | 1199 | ||
| 25 | Ideogram 4.5ideogram/ideogram-4.5 | Ideogram | 60.8 | 2 of 3 benchmarks | 1346 | 1333 | |
| 26 | Mai Image 2microsoft/mai-image-2 | Microsoft | 60.5 | 1 of 3 benchmarks | 1183 | ||
| 27 | Qwen Image 2.0 Proqwen/qwen-image-2.0-pro | Qwen | 60.2 | 2 of 3 benchmarks | 1190 | 1303 | |
| 28 | Grok Imagine Imagex-ai/grok-imagine-image | SpaceXAI | 60.0 | 2 of 3 benchmarks | 1168 | 1328 | |
| 29 | Recraft V4.1 Utility Prorecraft/recraft-v4.1-utility-pro | Recraft | 59.7 | 1 of 3 benchmarks | 1169 | ||
| 30 | Uni-1.1luma/uni-1.1 | Luma AI | 59.5 | 3 of 3 benchmarks | 1182 | 1316 | 1287 |
| 31 | Cosmos3 Super Text2imagenvidia/cosmos3-super-text2image | NVIDIA | 58.8 | 1 of 3 benchmarks | 1164 | ||
| 32 | Grok Imagine Image Prox-ai/grok-imagine-image-pro | SpaceXAI | 58.0 | 1 of 3 benchmarks | 1160 | ||
| 33 | Reve v1.5reve/reve-v1.5 | Reve | 57.1 | 1 of 3 benchmarks | 1154 | ||
| 34 | Seedream 4.5bytedance-seed/seedream-4.5 | ByteDance Seed | 57.0 | 3 of 3 benchmarks | 1148 | 1303 | 1290 |
| 35 | Hunyuan Image 3.0tencent/hunyuan-image-3.0 | Tencent | 56.3 | 1 of 3 benchmarks | 1152 | ||
| 36 | Imagen Ultra 4.0 Generate 001google/imagen-ultra-4.0-generate-001 | 55.0 | 1 of 3 benchmarks | 1146 | |||
| 37 | Wan2.6 T2Ialibaba/wan2.6-t2i | Alibaba | 53.8 | 1 of 3 benchmarks | 1138 | ||
| 38 | Hunyuan Image 3.0 Instructtencent/hunyuan-image-3.0-instruct | Tencent | 53.7 | 1 of 3 benchmarks | 1301 | ||
| 39 | Recraft V4.1 Prorecraft/recraft-v4.1-pro | Recraft | 52.9 | 1 of 3 benchmarks | 1132 | ||
| 40 | FLUX.2 Maxblack-forest-labs/flux.2-max | Black Forest Labs | 52.8 | 3 of 3 benchmarks | 1161 | 1262 | 1250 |
| 41 | Seedream 5.0 Litebytedance-seed/seedream-5-0-lite | ByteDance Seed | 52.6 | 3 of 3 benchmarks | 1141 | 1293 | 1266 |
| 42 | Krea 2 Mediumkrea/krea-2-medium | Krea | 52.5 | 1 of 3 benchmarks | 1128 | ||
| 43 | Imagen 4.0 Generate 001google/imagen-4.0-generate-001 | 52.1 | 1 of 3 benchmarks | 1128 | |||
| 44 | Qwen Image 2512qwen/qwen-image-2512 | Qwen | 51.6 | 1 of 3 benchmarks | 1124 | ||
| 45 | Wan2.7 Image Proalibaba/wan2.7-image-pro | Alibaba | 51.4 | 3 of 3 benchmarks | 1103 | 1302 | 1290 |
| 46 | Recraft V4recraft/recraft-v4 | Recraft | 50.8 | 1 of 3 benchmarks | 1116 | ||
| 47 | Nano Banana (Gemini 2.5 Flash Image)google/gemini-2.5-flash-image | 50.8 | 3 of 3 benchmarks | 1148 | 1292 | 1234 | |
| 48 | FLUX.2 Problack-forest-labs/flux.2-pro | Black Forest Labs | 50.2 | 3 of 3 benchmarks | 1153 | 1248 | 1240 |
| 49 | Reve v1.1reve/reve-v1.1 | Reve | 50.2 | 2 of 3 benchmarks | 1261 | 1261 | |
| 50 | Krea 2 Turbokrea/krea-2-turbo | Krea | 50.0 | 1 of 3 benchmarks | 1116 | ||
| 51 | Kling Image O1kwaivgi/kling-image-o1 | Kling AI | 49.2 | 2 of 3 benchmarks | 1250 | 1253 | |
| 52 | Wan2.7 Imagealibaba/wan2.7-image | Alibaba | 49.1 | 3 of 3 benchmarks | 1100 | 1300 | 1280 |
| 53 | Hidream O1 Imagehidream/hidream-o1-image | HiDream | 49.1 | 1 of 3 benchmarks | 1115 | ||
| 54 | Krea 2 Largekrea/krea-2-large | Krea | 48.7 | 1 of 3 benchmarks | 1114 | ||
| 55 | Qwen Image Editqwen/qwen-image-edit | Qwen | 48.5 | 1 of 3 benchmarks | 1241 | ||
| 56 | FLUX.2 Flexblack-forest-labs/flux.2-flex | Black Forest Labs | 48.0 | 3 of 3 benchmarks | 1156 | 1224 | 1234 |
| 57 | Recraft V4.1 Flashrecraft/recraft-v4.1-flash | Recraft | 46.6 | 1 of 3 benchmarks | 1098 | ||
| 58 | Mai Image 1microsoft/mai-image-1 | Microsoft | 46.2 | 1 of 3 benchmarks | 1093 | ||
| 59 | Seedream 4 (2K)bytedance/seedream-4:2k | ByteDance | 46.0 | 3 of 3 benchmarks | 1138 | 1267 | 1198 |
| 60 | Z Image Turboalibaba/z-image-turbo | Alibaba | 45.7 | 1 of 3 benchmarks | 1082 | ||
| 61 | Seedream 3bytedance/seedream-3 | ByteDance | 45.3 | 1 of 3 benchmarks | 1081 | ||
| 62 | Reve v1reve/reve-v1 | Reve | 45.1 | 2 of 3 benchmarks | 1235 | 1224 | |
| 63 | Qwen Image Prompt Extendqwen/qwen-image-prompt-extend | Qwen | 44.0 | 1 of 3 benchmarks | 1061 | ||
| 64 | Wan2.6 Imagealibaba/wan2.6-image | Alibaba | 43.6 | 2 of 3 benchmarks | 1231 | 1220 | |
| 65 | Flux 2 Devblack-forest-labs/flux-2-dev | Black Forest Labs | 43.4 | 3 of 3 benchmarks | 1145 | 1224 | 1202 |
| 66 | Reve v1.1 Fastreve/reve-v1.1-fast | Reve | 43.4 | 1 of 3 benchmarks | 1207 | ||
| 67 | Qwen Imageqwen/qwen-image | Qwen | 43.2 | 1 of 3 benchmarks | 1057 | ||
| 68 | Imagen 3.0 Generate 002google/imagen-3.0-generate-002 | 42.8 | 1 of 3 benchmarks | 1057 | |||
| 69 | Reve Edit Fastreve/reve-edit-fast | Reve | 42.8 | 1 of 3 benchmarks | 1198 | ||
| 70 | Ideogram v3 Qualityideogram/ideogram-v3-quality | Ideogram | 42.4 | 1 of 3 benchmarks | 1049 | ||
| 71 | Photonluma/photon | Luma AI | 41.9 | 1 of 3 benchmarks | 1036 | ||
| 72 | Qwen Image Edit 2511qwen/qwen-image-edit-2511 | Qwen | 41.2 | 2 of 3 benchmarks | 1233 | 1173 | |
| 73 | Runway Gen4runway/runway-gen4 | Runway | 41.1 | 1 of 3 benchmarks | 1024 | ||
| 74 | Recraft V3recraft/recraft-v3 | Recraft | 40.7 | 1 of 3 benchmarks | 1022 | ||
| 75 | Flux 1.1 Problack-forest-labs/flux-1.1-pro | Black Forest Labs | 40.3 | 1 of 3 benchmarks | 1016 | ||
| 76 | Ideogram v2ideogram/ideogram-v2 | Ideogram | 39.8 | 1 of 3 benchmarks | 1013 | ||
| 77 | Seedream 4 High Res Falbytedance/seedream-4-high-res-fal | ByteDance | 39.7 | 3 of 3 benchmarks | 1113 | 1216 | 1209 |
| 78 | Lucid Originleonardo/lucid-origin | Leonardo AI | 39.4 | 1 of 3 benchmarks | 1013 | ||
| 79 | GLM Imagez-ai/glm-image | Z.ai | 39.0 | 1 of 3 benchmarks | 1012 | ||
| 80 | Flux 2 Klein 9Bblack-forest-labs/flux-2-klein-9b | Black Forest Labs | 38.9 | 3 of 3 benchmarks | 1070 | 1224 | 1211 |
| 81 | Flux 1 Devblack-forest-labs/flux-1-dev | Black Forest Labs | 38.1 | 1 of 3 benchmarks | 969 | ||
| 82 | Seedream 4 Falbytedance/seedream-4-fal | ByteDance | 37.9 | 3 of 3 benchmarks | 1116 | 1209 | 1158 |
| 83 | DALL E 3openai/dall-e-3 | OpenAI | 37.7 | 1 of 3 benchmarks | 968 | ||
| 84 | Wan2.5 I2I Previewalibaba/wan2.5-i2i-preview | Alibaba | 37.1 | 3 of 3 benchmarks | 1117 | 1180 | 1165 |
| 85 | Stable Diffusion v35 Largestability-ai/stable-diffusion-v35-large | Stability AI | 36.9 | 1 of 3 benchmarks | 938 | ||
| 86 | Step1x Editstepfun/step1x-edit | StepFun | 36.5 | 1 of 3 benchmarks | 996 | ||
| 87 | GPT Image 1openai/gpt-image-1 | OpenAI | 33.8 | 3 of 3 benchmarks | 1115 | 1139 | 1126 |
| 88 | FLUX.2 Klein 4Bblack-forest-labs/flux.2-klein-4b | Black Forest Labs | 32.5 | 3 of 3 benchmarks | 1030 | 1187 | 1166 |
| 89 | GPT Image 1 Miniopenai/gpt-image-1-mini | OpenAI | 32.0 | 3 of 3 benchmarks | 1109 | 1124 | 1111 |
| 90 | Flux 1 Kontext (max)black-forest-labs/flux-1-kontext:max | Black Forest Labs | 31.4 | 3 of 3 benchmarks | 1075 | 1182 | 1062 |
| 91 | Flux 1 Kontext Problack-forest-labs/flux-1-kontext-pro | Black Forest Labs | 30.4 | 3 of 3 benchmarks | 1059 | 1177 | 1063 |
| 92 | Seededit 3.0bytedance/seededit-3.0 | ByteDance | 30.1 | 2 of 3 benchmarks | 1139 | 1042 | |
| 93 | Bagelbytedance/bagel | ByteDance | 27.8 | 2 of 3 benchmarks | 897 | 1027 | |
| 94 | Gemini 2.0 Flash Preview Image Generationgoogle/gemini-2.0-flash-preview-image-generation | 24.7 | 3 of 3 benchmarks | 976 | 1082 | 1058 | |
| 95 | Flux 1 Kontext Devblack-forest-labs/flux-1-kontext-dev | Black Forest Labs | 24.4 | 3 of 3 benchmarks | 941 | 1150 | 1036 |
How this ranks
Every benchmark value becomes a percentile among the models that have it, so accuracy scores, Elo ratings and word error rates compare without hand-tuned scaling. Metrics where lower is better are inverted first. Raw values are never summed or averaged across benchmarks. A model's mean percentile is then shrunk toward the mean of the models that were broadly benchmarked, so a model tested twice cannot outrank a broadly tested one on two lucky results. Turning a data source off runs that same ranking code again in your browser over the sources you left on.
A model scored on fewer than 2 of the 3 ranked benchmarks in this category still ranks here, on the benchmarks it does have, and its row carries a partial coverage mark. On an equal score it sits under the model that earned the same number across more of the board.
Data sources
Turn a source off to drop every benchmark it feeds and re-rank the board in your browser. Your choice follows you across the leaderboard pages.
- LMArenaCC BY 4.0
Arena ratings by LMArena, from the public leaderboard dataset.