Qwen Releases Qwen-Image-2.1: A Powerful Open Model for Image Generation and Editing
Qwen has launched Qwen-Image-2.1, a 7B-parameter open model for image generation and editing that outperforms Nano Banana 2.0.
Qwen has released another powerful update. Introducing Qwen-Image-2.1 — a 7-billion parameter open model for image generation and editing, which outperforms Nano Banana 2.0.
The model supports up to 10 reference images, local editing, and native transparency. This allows users to generate images with transparent backgrounds immediately and cut out objects without needing a separate mask.
You can access the model on Hugging Face: Qwen/Qwen-Image-2.1
Additionally, the community has already created an uncensored version: Qwen-Image-2.1-Uncensored-GGUF
Why it matters
AnalysisThis release provides a robust open-source alternative for image generation, offering advanced features like native transparency support that streamline creative workflows.
Discuss in community
Share your questions and insights with developers
Related in Ecosystem
Practical guides, core terminology, and agent skills
GPT Image / DALL-E (Image Generation in ChatGPT)
An integrated visual content generation tool directly within the ChatGPT dialogue. It allows for the creation of illustrations, concept art, poster texts, and local editing of image fragments.
Negative Prompts in Image Generation
A control mechanism in diffusion models (Stable Diffusion, Midjourney --no) that directs the mathematical vector of noise reduction away from unwanted images. It filters artifacts such as extra limbs, blurriness, watermarks, and cartoonish effects.
Thinking Machines Releases Inkling 2 Small: a 594B MoE Model That Can Fine-Tune Other Models
Thinking Machines released Inkling 2 Small, a 594B-parameter MoE model that leads among open models in several benchmarks and demonstrated autonomous fine-tuning of Qwen3.5-4B.
Jev and System One AI: A Complete Practical Guide to the Decision Model from TypeSafe
A deep engineering breakdown of the flagship Jev model: System One architecture versus LLMs, Choice/Score/Noul primitives, request parallelism, confidence mathematics, video use cases, code, and an overview of local open-weight alternatives (Laya, GLiNER2.5, CLM-8B).