• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
Technology

Qwen releases Image 2.1 with transparent editing — under a research-only license

Qwen says its new image model can generate and edit transparent layers using up to 10 reference images. The downloadable weights do not carry commercial-use rights.

PrathamPR
VORTEX: AI Bros & AI Arena, Peak AI BuzzVA
TechmemeTE
4 Sources, 20d ago, first seen 20d ago

TLDR

Qwen has released Image 2.1, which it says combines image generation and editing, handles transparent RGBA images, and accepts up to 10 reference images. The 7-billion-parameter figure applies to its visual generation component. Although the weights are available to download, the Qwen Research License permits only non-commercial research and evaluation; commercial use requires a separate license.

Combined views

66.1K

4 Sources, first seen 20d ago

264 likes27 comments48 saves181 reposts

Combined views

66.1K

4 Sources, first seen 20d ago

264 likes27 comments48 saves181 reposts
Qwen logo
Image: Qwen / Alibaba Cloud, Apache License 2.0, via Wikimedia Commons

Qwen has released Qwen-Image-2.1, a model for generating and editing images. Its model card describes a 7-billion-parameter visual generation component. The model's stated capabilities include creating images with transparent backgrounds, editing transparent layers, and using as many as 10 reference images to guide an edit.

Featured Source

Those features could be useful for compositing and targeted changes: Qwen says users can indicate an area to edit with a circle, painted annotation, or mask. The card includes instructions for running generation and editing through Diffusers, making the release more than an announcement of a hosted demo. The quality and speed claims in Qwen's materials remain the developer's claims, not independently verified comparisons.

The availability of the weights comes with an important limit. Qwen labels the release “open-source,” but the linked Qwen Research License allows use of the materials only for non-commercial research or evaluation. Anyone wanting to use them commercially must obtain a separate license. “Open-weight” describes the download; it should not be mistaken for unrestricted commercial permission.

Sentiment

Positive——Negative

Summary

Not enough discussion yet.

No sentiment analysis available yet.

Sentiment

Positive——Negative

Summary

Not enough discussion yet.

No sentiment analysis available yet.

Related

Alibaba stock reportedly fell after earnings as profit dropped 76% and capex rose 75%

Tech in Asia reported that Alibaba’s profit fell 76% while capex rose 75%, and said the stock dropped after earnings as investors may be missing its broader chips, cloud, AI and apps push.

China weighs allowing Nvidia RTX Pro 5500 purchases by Alibaba and ByteDance

The Information reports that officials asked companies how many chips they want and how they plan to use them, but approval terms remain unclear.

Alibaba unveils Zhenwu V900 AI chip, targets more than 20GW of global data center capacity by 2032

A post says the V900 is designed to train and run AI models, with mass production and commercial release expected in Q1 2027. It describes claimed performance roughly 3x higher than the chip’s predecessor.

4 Sources

Pratham@PrathkumGoogle should be nervous. Qwen's open source model just beat Nano Banana 2.0. 7B image params, generation and editing in one model, native transparent PNGs, up to 10 reference images.20d
VORTEX: AI Bros & AI Arena, Peak AI Buzz@VORTEX_Promos🚨 Qwen-Image 2.1 #Alibaba just dropped #Qwen-Image 2.1 — a unified text-to-image + image editing model with only 7B parameters. 🔥 Native 2K 🔥 Up to 10 reference images 🔥 Better identity & consistency 🔥 Masks/scribbles for local editing 🔥 Native RGBA + transparency 🔥 ComfyUI support from day one The really interesting part: RGBA editing and object extraction from regular RGB images. Need to test this one. 💾 VRAM-wise, 12 GB looks like a realistic target with quantization/offloading. 16 GB should be very comfortable. The previous Qwen Image was ~20B. This one is just 7B. #ai 🔗 https://modelscope.ai/models/Qwen/Qwen-Image-2.1/summary20d
Techmeme@TechmemeAlibaba releases Qwen-Image-2.1, a 7B open-weight model it says outperforms most closed-source models, with native transparency and up to ten reference images (Qwen) (Visit Techmeme dot com for the link and full context!)20d
Sree@sreextsQwen-Image-2.1 dropped open-source — 7B backbone, gen + edit in one model, up to 10 reference images with native RGBA and transparent PNGs. Day-0 ComfyUI + Diffusers + MLX, Q4_K_M fits in 4.3GB on your laptop. Closed labs are still trying to gate "edit with reference images" behind $50/mo plans, where as @Alibaba_Qwen just put it on HF for free20d
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    Qwen-Image-2.1Alibaba Group

    4 Sources

    Pratham@PrathkumGoogle should be nervous. Qwen's open source model just beat Nano Banana 2.0. 7B image params, generation and editing in one model, native transparent PNGs, up to 10 reference images.20d
    VORTEX: AI Bros & AI Arena, Peak AI Buzz@VORTEX_Promos🚨 Qwen-Image 2.1 #Alibaba just dropped #Qwen-Image 2.1 — a unified text-to-image + image editing model with only 7B parameters. 🔥 Native 2K 🔥 Up to 10 reference images 🔥 Better identity & consistency 🔥 Masks/scribbles for local editing 🔥 Native RGBA + transparency 🔥 ComfyUI support from day one The really interesting part: RGBA editing and object extraction from regular RGB images. Need to test this one. 💾 VRAM-wise, 12 GB looks like a realistic target with quantization/offloading. 16 GB should be very comfortable. The previous Qwen Image was ~20B. This one is just 7B. #ai 🔗 https://modelscope.ai/models/Qwen/Qwen-Image-2.1/summary20d
    Techmeme@TechmemeAlibaba releases Qwen-Image-2.1, a 7B open-weight model it says outperforms most closed-source models, with native transparency and up to ten reference images (Qwen) (Visit Techmeme dot com for the link and full context!)20d
    Sree@sreextsQwen-Image-2.1 dropped open-source — 7B backbone, gen + edit in one model, up to 10 reference images with native RGBA and transparent PNGs. Day-0 ComfyUI + Diffusers + MLX, Q4_K_M fits in 4.3GB on your laptop. Closed labs are still trying to gate "edit with reference images" behind $50/mo plans, where as @Alibaba_Qwen just put it on HF for free20d
    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet