Thanks @arena for the recognition! 🏆 Qwen-Image-2.1 is now the #1 open-source model in both the Image Edit and Text-to-Image Arenas. Try it now and show us what you create! 🎨
Arena.ai
@arena
Sep 22
Qwen-Image-2.1 by @Alibaba_Qwen just landed as the #1 open source model in the Image Edit Arena and Text-to-Image Arena! With 1367 pts in the Image Edit Arena, Qwen-Image-2.1 took the #1 spot among open. It landed #16 overall, just 3 pts from GPT-Image-1.5-high-fidelity at #15. See the leaderboard for the Text-to-Image arena below. Congrats to the @Alibaba_Qwen team on this contribution to the open source ecosystem!

Sep 23, 2026 · 1:24 AM UTC

33
27
546
56,636
Sort replies: Relevant Recent Liked
Replying to @Alibaba_Qwen @arena
Is the photo that I am making with this model free to use and sell?
235
Replying to @Alibaba_Qwen @arena
Does it handle complex lighting edits well?
244
Replying to @Alibaba_Qwen @arena
if youre using Loras put your CFG to 3 not 1. Youre welcome
245
Replying to @Alibaba_Qwen @arena
Did anyone able to use it via @UnslothAI studio?
320
Replying to @Alibaba_Qwen @arena
A lot more organic compared to previous models especially on the art side. Thanks for sharing it with the World!
287
Replying to @Alibaba_Qwen @arena
77
Replying to @Alibaba_Qwen @arena
Image editing is the real bottleneck for robotics sim data. If it handles complex instruction following well, that changes how we generate training sets for vision tasks. Huge win for open source.
53
Replying to @Alibaba_Qwen @arena
a model that edits reality and generates new problems for group chats
122
Replying to @Alibaba_Qwen @arena
Nice work. Magica already routes to whichever image model tops the arena, so this win just landed in users' hands automatically.
53
Replying to @Alibaba_Qwen @arena
text-to-image being competitive was expected, but landing 3 pts behind GPT-Image-1.5 in Image Edit is the real signal. open models usually fall off hard once you ask for multi-turn edits or reference preservation. 1367 on blind human preference closes that gap fast.
208
Replying to @Alibaba_Qwen @arena
Arena 分高是好事，但 Image Edit 最怕 leakage 和 prompt 套路过拟合。有没有开放一组「真实生产编辑」样本（Logo 擦除、文字重排、局部材质替换）的前后对比和失败案例？否则 #1 open 很难直接翻译成可上线质量。
79
Replying to @Alibaba_Qwen @arena
open just took both image arenas. closed ops keep paying for a shrinking gap
167
Replying to @Alibaba_Qwen @arena
Dominating both generation and editing at the same time is massive for open weights, well deserved!
107
Replying to @Alibaba_Qwen @arena
Wow, the open source one leading the Image Edit arena! Thank @Alibaba_Qwen
91
Replying to @Alibaba_Qwen @arena
Any word on a commercial license or is it research only for good?
85
Replying to @Alibaba_Qwen @arena
congrats! excited to get image 2.1 integrated into my ecosystem!
124
Replying to @Alibaba_Qwen @arena
Open-source image leadership is increasingly measurable through separate edit and generation arenas.
120
Replying to @Alibaba_Qwen @arena
the worst model. shouldn't be on the chart
8
Replying to @Alibaba_Qwen @arena
#1 looks good on Qwen. Congrats! 🏆
119
Replying to @Alibaba_Qwen @arena
更强大，真的很强
49
Replying to @Alibaba_Qwen @arena
Open models getting genuinely competitive changes the economics of AI.
49
Replying to @Alibaba_Qwen @arena
Arena crown secured, now let's see what the artists actually ship with it.
112
Replying to @Alibaba_Qwen @arena
open source model? what the fuck
16
Replying to @Alibaba_Qwen @arena
Upload a tweet about EVA
73