Best AI for multi-image editing: AI Image Arena ranked
Nov 17, 2026 · 8,873,388 votes · 48 models
GPT-Image-2.5 Sunburst by OpenAI leads with 1527 points, ahead of GPT-Image-2.5 Flare (1475) and MAI-Image-2.7 (1458).
| Rank | Model | Score | Spread | Votes |
|---|---|---|---|---|
| 1 | GPT-Image-2.5 SunburstOpenAI · Proprietary | 1527 ±5 | 1 | 70,812 |
| 2 | GPT-Image-2.5 FlareOpenAI · Proprietary | 1475 ±5 | 2 | 69,942 |
| 3 | MAI-Image-2.7Microsoft AI · Proprietary | 1458 ±10 | 3-5 | 3,647 |
| 4 | GPT-Image-2 (Medium)OpenAI · Proprietary | 1455 ±3 | 3-4 | 167,372 |
| 5 | GPT-Image-2.6 Flare (Medium)OpenAI · Proprietary | 1442 ±7 | 4-7 | 10,635 |
| 6 | Gemini Nano Banana 2.1Google · Proprietary | 1432 ±7 | 5-7 | 9,631 |
| 7 | Krea 3 UltraKrea · Proprietary | 1431 ±7 | 5-7 | 11,506 |
| 8 | Seedream 5.0 ProByteDance · Proprietary | 1412 ±4 | 8 | 119,364 |
| 9 | Muse ImageMeta · Proprietary | 1400 ±4 | 9 | 129,822 |
| 10 | Gemini 3.1 Flash Image (nano-banana-2, Web Search)Google · Proprietary | 1372 ±3 | 10-12 | 146,492 |
| 11 | MiniMax H3-ImageMiniMax · minimax-h3-image-community-license-agreement | 1367 ±8 | 10-13 | 6,296 |
| 12 | Gemini 3 Pro Image Preview (nano-banana-pro)Google · Proprietary | 1366 ±4 | 10-13 | 293,627 |
| 13 | Gemini 3 Pro Image (nano-banana-pro, 2K)Google · Proprietary | 1364 ±3 | 11-13 | 278,901 |
| 14 | ChatGPT Image Latest (High Fidelity, 20251216)OpenAI · Proprietary | 1350 ±3 | 14-15 | 278,465 |
| 15 | Reve 2.0Reve · Proprietary | 1346 ±7 | 14-17 | 9,927 |
| 16 | GPT-Image-1.5 (High Fidelity)OpenAI · Proprietary | 1342 ±3 | 15-17 | 245,850 |
| 17 | Ideogram 4.5Ideogram · Proprietary | 1335 ±5 | 15-17 | 26,104 |
| 18 | UNI-1.1 MaxLuma AI · Proprietary | 1308 ±5 | 18 | 21,241 |
| 19 | Seedream 4.5ByteDance · Proprietary | 1290 ±3 | 19-22 | 621,695 |
| 20 | Gemini 3.1 Flash-Lite Image (nano-banana-2-lite)Google · Proprietary | 1290 ±5 | 19-22 | 28,298 |
| 21 | Wan 2.7 Image ProAlibaba · Proprietary | 1290 ±5 | 19-23 | 17,515 |
| 22 | UNI-1.1Luma AI · Proprietary | 1288 ±4 | 19-23 | 114,816 |
| 23 | Wan 2.7 ImageAlibaba · Proprietary | 1280 ±5 | 21-23 | 18,777 |
| 24 | Seedream 5.0 LiteByteDance · Proprietary | 1266 ±3 | 24-25 | 405,500 |
| 25 | Reve v1.1Reve · Proprietary | 1262 ±4 | 24-26 | 264,754 |
| 26 | Kling Image O1Kling AI · Proprietary | 1253 ±5 | 25-27 | 47,138 |
| 27 | FLUX.2 [max]Black Forest Labs · Proprietary | 1250 ±4 | 26-27 | 142,261 |
| 28 | FLUX.2 [pro]Black Forest Labs · Proprietary | 1240 ±3 | 28-29 | 265,044 |
| 29 | FLUX.2 [flex]Black Forest Labs · Proprietary | 1235 ±4 | 28-31 | 131,132 |
| 30 | Gemini 2.5 Flash Image Preview (nano-banana)Google · Proprietary | 1234 ±2 | 29-31 | 4,704,334 |
| 31 | Reve v1Reve · Proprietary | 1224 ±9 | 29-33 | 547,185 |
| 32 | Wan 2.6 ImageAlibaba · Proprietary | 1221 ±3 | 31-32 | 240,762 |
| 33 | FLUX.2 [klein] 9BBlack Forest Labs · flux-non-commercial-license | 1212 ±4 | 32-35 | 155,186 |
| 34 | Seedream 4.0 (High Res, fal)ByteDance · Proprietary | 1209 ±4 | 33-36 | 1,039,361 |
| 35 | FLUX.2 [dev]Black Forest Labs · flux-non-commercial-license | 1202 ±5 | 34-36 | 118,265 |
| 36 | Seedream 4.0 (2K)ByteDance · Proprietary | 1198 ±11 | 33-36 | 30,030 |
| 37 | Qwen-Image-Edit (2511)Alibaba · Apache 2.0 | 1174 ±4 | 37-38 | 228,772 |
| 38 | FLUX.2 [klein] 4BBlack Forest Labs · Apache 2.0 | 1166 ±4 | 37-40 | 154,545 |
| 39 | Wan 2.5 PreviewAlibaba · Proprietary | 1165 ±4 | 38-40 | 123,175 |
| 40 | Seedream 4.0 (fal)ByteDance · Proprietary | 1158 ±6 | 38-40 | 305,048 |
| 41 | P-Image-EditUnknown lab · Proprietary | 1146 ±4 | 41 | 132,837 |
| 42 | GPT-Image-1OpenAI · Proprietary | 1126 ±4 | 42 | 1,825,646 |
| 43 | GPT-Image-1 miniOpenAI · Proprietary | 1111 ±4 | 43 | 631,222 |
| 44 | FLUX.1 Kontext [pro]Black Forest Labs · Proprietary | 1063 ±12 | 44-47 | 140,754 |
| 45 | FLUX.1 Kontext [max]Black Forest Labs · Proprietary | 1062 ±9 | 44-47 | 51,921 |
| 46 | Gemini 2.0 Flash Preview Image GenerationGoogle · Proprietary | 1058 ±6 | 44-47 | 3,145,326 |
| 47 | SeedEdit 3.0ByteDance · Proprietary | 1042 ±16 | 44-48 | 86,777 |
| 48 | FLUX.1 Kontext [dev]Black Forest Labs · flux-1-dev-non-commercial-license | 1036 ±10 | 47-48 | 129,066 |
Score: Bradley-Terry rating - the bar shows it on one scale for the whole board, the thin line is the 95% confidence interval (± half of it next to the number). Spread: every rank the model could hold given that interval - models whose lines overlap are statistically tied.
Data: LMArena (lmarena-ai/leaderboard-dataset), CC BY 4.0. Published Nov 17, 2026. Flashlook mirrors the rankings unchanged - about the data.
Multi-image editing combines several inputs - for example a model reference and a garment photo - into one image. It's the closest public benchmark to virtual on-model photography.
Shoot your products on-modelConfidence intervals on model strength
Dot = score, line = 95% interval. Hover a dot for the numbers.
Top 40 shown; the table lists all 48.

