Alibaba's open-weight Qwen-Image-2.1 claims to beat closed models in image generation with just 7 billion parameters
Alibaba has released Qwen-Image-2.1, an open-weight diffusion transformer model with 7 billion parameters capable of generating and editing images. The model supports native RGBA transparency and can process up to ten reference images simultaneously, allowing it to perform complex tasks on consumer-grade hardware. While the weights are publicly available, the research license prohibits commercial use without a separate agreement. This release is significant because it demonstrates high-level image generation performance from a compact model, potentially shifting industry expectations for local AI capabilities.
Covered by 13 sources
- TThe Decoder↗Matthias Bastian2d ago
- MMarkTechPost↗Michal Sutter1d ago
- Ffinance.biggo.com↗2d ago
- TTechNode↗2d ago
- FFoneArena.com↗2d ago
- MMIXED Reality News↗1d ago
- KKuCoin↗2d ago
- EeWeek↗1d ago
- Eeu.36kr.com↗2d ago
- PPandaily↗2d ago
- Mmarktechpost.com↗1d ago
- Ttech-insider.org↗2d ago
- Tthe-decoder.com↗2d ago