---
title: "Best AI for image generation"
description: "Image models ranked on the LMArena text to image, image edit and multi-image edit arenas, with coverage shown for every row."
url: "https://aldena.ai/best-ai-for-image-generation"
---

# Best AI for image generation

Human preference across generation and editing, rated blind, pairwise.

Aldena runs these models inside your team rooms. [See what each one costs](https://aldena.ai/models).

| rank | model | vendor | score | coverage (benchmarks) | LMArena text to image (elo) | LMArena image edit (elo) | LMArena multi-image edit (elo) |
| --- | --- | --- | --- | --- | --- | --- | --- |
| 1 | GPT Image 2 (medium) | OpenAI | 81.6 | 3 of 3 | 1380 | 1463 | 1454 |
| 2 | Muse Image | Meta | 79.6 | 3 of 3 | 1283 | 1405 | 1401 |
| 3 | Seedream 5.0 Pro | ByteDance Seed | 78.2 | 3 of 3 | 1257 | 1393 | 1414 |
| 4 | Nano Banana 2 (Gemini 3.1 Flash Image) | Google | 75.1 | 3 of 3 | 1263 | 1385 | 1365 |
| 5 | Nano Banana Pro (Gemini 3 Pro Image Preview) | Google | 74.1 | 3 of 3 | 1232 | 1385 | 1368 |
| 6 | Nano Banana Pro (Gemini 3 Pro Image) (2K) | Google | 74.0 | 3 of 3 | 1246 | 1389 | 1364 |
| 7 | MAI-Image-2.5 | Microsoft | 73.5 | 2 of 3 | 1256 | 1402 | — |
| 8 | Reve 2.0 | Reve | 72.6 | 3 of 3 | 1270 | 1358 | 1344 |
| 9 | Reve 2.1 | Reve | 72.0 | 2 of 3 | 1302 | 1374 | — |
| 10 | Chatgpt Image High Fidelity | OpenAI | 71.0 | 2 of 3 | — | 1390 | 1353 |
| 11 | GPT Image 1.5 High Fidelity | OpenAI | 70.5 | 3 of 3 | 1239 | 1370 | 1342 |
| 12 | Grok Imagine Image Quality | SpaceXAI | 70.3 | 2 of 3 | 1228 | 1390 | — |
| 13 | Uni-1.1 (max) | Luma AI | 67.8 | 3 of 3 | 1188 | 1334 | 1308 |
| 14 | Qwen Image 3.0 Pro (partial coverage) | Qwen | 67.0 | 1 of 3 | 1258 | — | — |
| 15 | Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image) | Google | 66.8 | 3 of 3 | 1251 | 1314 | 1286 |
| 16 | Uni-1.1 | Luma AI | 64.1 | 3 of 3 | 1180 | 1312 | 1287 |
| 17 | Qwen Image 2.0 Pro | Qwen | 64.0 | 2 of 3 | 1191 | 1303 | — |
| 18 | Grok Imagine Image | xAI | 63.8 | 2 of 3 | 1172 | 1330 | — |
| 19 | Ideogram 4.0 Quality (partial coverage) | Ideogram | 63.3 | 1 of 3 | 1206 | — | — |
| 20 | Mai Image 2 (partial coverage) | Microsoft | 61.9 | 1 of 3 | 1182 | — | — |
| 21 | Cosmos3 Super Text2image (partial coverage) | NVIDIA | 61.5 | 1 of 3 | 1181 | — | — |
| 22 | Seedream 4.5 | ByteDance Seed | 60.1 | 3 of 3 | 1146 | 1301 | 1291 |
| 23 | Recraft V4.1 Utility Pro (partial coverage) | Recraft | 60.1 | 1 of 3 | 1169 | — | — |
| 24 | Grok Imagine Image Pro (partial coverage) | xAI | 59.1 | 1 of 3 | 1161 | — | — |
| 25 | Hunyuan Image 3.0 Instruct (partial coverage) | Tencent | 57.8 | 1 of 3 | — | 1302 | — |
| 26 | Reve v1.5 (partial coverage) | Reve | 57.8 | 1 of 3 | 1154 | — | — |
| 27 | Hunyuan Image 3.0 (partial coverage) | Tencent | 57.3 | 1 of 3 | 1151 | — | — |
| 28 | FLUX.2 Max | Black Forest Labs | 56.5 | 3 of 3 | 1162 | 1262 | 1250 |
| 29 | Imagen Ultra 4.0 Generate 001 (partial coverage) | Google | 56.4 | 1 of 3 | 1148 | — | — |
| 30 | Seedream 5.0 Lite | ByteDance Seed | 55.8 | 3 of 3 | 1136 | 1293 | 1270 |
| 31 | Wan2.7 Image Pro | Alibaba | 55.4 | 3 of 3 | 1103 | 1302 | 1289 |
| 32 | Nano Banana (Gemini 2.5 Flash Image) | Google | 55.0 | 3 of 3 | 1150 | 1294 | 1234 |
| 33 | Wan2.6 T2I (partial coverage) | Alibaba | 54.0 | 1 of 3 | 1136 | — | — |
| 34 | FLUX.2 Pro | Black Forest Labs | 53.9 | 3 of 3 | 1155 | 1244 | 1238 |
| 35 | Reve v1.1 | Reve | 53.7 | 2 of 3 | — | 1261 | 1261 |
| 36 | Recraft V4.1 Pro (partial coverage) | Recraft | 53.6 | 1 of 3 | 1130 | — | — |
| 37 | Wan2.7 Image | Alibaba | 53.1 | 3 of 3 | 1100 | 1301 | 1280 |
| 38 | Imagen 4.0 Generate 001 (partial coverage) | Google | 53.1 | 1 of 3 | 1129 | — | — |
| 39 | Qwen Image 2512 (partial coverage) | Qwen | 52.6 | 1 of 3 | 1126 | — | — |
| 40 | Kling Image O1 | Kling AI | 52.5 | 2 of 3 | — | 1251 | 1252 |
| 41 | Krea 2 (medium) (partial coverage) | Krea | 52.2 | 1 of 3 | 1122 | — | — |
| 42 | Hidream O1 Image (partial coverage) | HiDream | 51.7 | 1 of 3 | 1118 | — | — |
| 43 | Wan2.5 T2I Preview (partial coverage) | Alibaba | 51.3 | 1 of 3 | 1117 | — | — |
| 44 | FLUX.2 Flex | Black Forest Labs | 51.1 | 3 of 3 | 1156 | 1225 | 1234 |
| 45 | Qwen Image Edit (partial coverage) | Qwen | 50.3 | 1 of 3 | — | 1241 | — |
| 46 | Recraft V4 (partial coverage) | Recraft | 49.9 | 1 of 3 | 1113 | — | — |
| 47 | Seedream 4 (2K) | ByteDance | 49.4 | 3 of 3 | 1140 | 1271 | 1201 |
| 48 | Krea 2 Turbo (partial coverage) | Krea | 48.9 | 1 of 3 | 1110 | — | — |
| 49 | Krea 2 Large (partial coverage) | Krea | 48.0 | 1 of 3 | 1105 | — | — |
| 50 | Reve v1 | Reve | 47.2 | 2 of 3 | — | 1234 | 1222 |
| 51 | Mai Image 1 (partial coverage) | Microsoft | 46.6 | 1 of 3 | 1093 | — | — |
| 52 | Seedream 3 (partial coverage) | ByteDance | 46.2 | 1 of 3 | 1082 | — | — |
| 53 | Wan2.6 Image | Alibaba | 46.0 | 2 of 3 | — | 1230 | 1219 |
| 54 | Z Image Turbo (partial coverage) | Alibaba | 45.7 | 1 of 3 | 1081 | — | — |
| 55 | Flux 2 Dev | Black Forest Labs | 45.7 | 3 of 3 | 1146 | 1224 | 1202 |
| 56 | Qwen Image Prompt Extend (partial coverage) | Qwen | 44.3 | 1 of 3 | 1060 | — | — |
| 57 | Reve v1.1 Fast (partial coverage) | Reve | 44.1 | 1 of 3 | — | 1207 | — |
| 58 | Qwen Image Edit 2511 | Qwen | 43.8 | 2 of 3 | — | 1235 | 1173 |
| 59 | Reve Edit Fast (partial coverage) | Reve | 43.5 | 1 of 3 | — | 1198 | — |
| 60 | Imagen 3.0 Generate 002 (partial coverage) | Google | 43.4 | 1 of 3 | 1058 | — | — |
| 61 | Qwen Image (partial coverage) | Qwen | 42.9 | 1 of 3 | 1057 | — | — |
| 62 | Ideogram v3 Quality (partial coverage) | Ideogram | 42.5 | 1 of 3 | 1049 | — | — |
| 63 | Seedream 4 High Res Fal | ByteDance | 42.2 | 3 of 3 | 1113 | 1217 | 1209 |
| 64 | Photon (partial coverage) | Luma AI | 42.0 | 1 of 3 | 1035 | — | — |
| 65 | Runway Gen4 (partial coverage) | Runway | 41.1 | 1 of 3 | 1025 | — | — |
| 66 | Flux 2 Klein 9B | Black Forest Labs | 40.8 | 3 of 3 | 1070 | 1224 | 1213 |
| 67 | Recraft V3 (partial coverage) | Recraft | 40.6 | 1 of 3 | 1021 | — | — |
| 68 | Flux 1.1 Pro (partial coverage) | Black Forest Labs | 40.2 | 1 of 3 | 1016 | — | — |
| 69 | Lucid Origin (partial coverage) | Leonardo AI | 39.7 | 1 of 3 | 1013 | — | — |
| 70 | Seedream 4 Fal | ByteDance | 39.5 | 3 of 3 | 1116 | 1210 | 1156 |
| 71 | Ideogram v2 (partial coverage) | Ideogram | 39.2 | 1 of 3 | 1013 | — | — |
| 72 | GLM Image (partial coverage) | Z.ai | 38.8 | 1 of 3 | 1010 | — | — |
| 73 | Flux 1 Dev (partial coverage) | Black Forest Labs | 37.8 | 1 of 3 | 969 | — | — |
| 74 | DALL E 3 (partial coverage) | OpenAI | 37.4 | 1 of 3 | 968 | — | — |
| 75 | Wan2.5 I2I Preview | Alibaba | 36.8 | 2 of 3 | — | 1181 | 1160 |
| 76 | Stable Diffusion v35 Large (partial coverage) | Stability AI | 36.4 | 1 of 3 | 938 | — | — |
| 77 | Step1x Edit (partial coverage) | StepFun | 36.0 | 1 of 3 | — | 998 | — |
| 78 | GPT Image 1 | OpenAI | 35.0 | 3 of 3 | 1115 | 1139 | 1126 |
| 79 | FLUX.2 Klein 4B | Black Forest Labs | 33.7 | 3 of 3 | 1030 | 1188 | 1166 |
| 80 | GPT Image 1 Mini | OpenAI | 33.0 | 3 of 3 | 1109 | 1124 | 1111 |
| 81 | Flux 1 Kontext (max) | Black Forest Labs | 32.0 | 3 of 3 | 1074 | 1181 | 1059 |
| 82 | Flux 1 Kontext Pro | Black Forest Labs | 31.3 | 3 of 3 | 1059 | 1176 | 1061 |
| 83 | Seededit 3.0 | ByteDance | 30.2 | 2 of 3 | — | 1139 | 1042 |
| 84 | Bagel | ByteDance | 27.5 | 2 of 3 | 898 | 1026 | — |
| 85 | Gemini 2.0 Flash Preview Image Generation | Google | 24.9 | 3 of 3 | 975 | 1081 | 1058 |
| 86 | Flux 1 Kontext Dev | Black Forest Labs | 24.6 | 3 of 3 | 940 | 1149 | 1034 |

## 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

- [LMArena](https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset): CC BY 4.0. Arena ratings by LMArena, from the public leaderboard dataset.