Alibaba open-sources Qwen-Image-2.1: 7B model with native RGBA transparency
AdinaYakup · x · 2026-09-21
Alibaba's Qwen team released Qwen-Image-2.1, a unified text-to-image generation and editing model with a 7B-parameter visual component (32 Single-Stream DiT layers). Highlights: native transparent (RGBA) image generation, support for up to 10 reference images with identity preservation for people and products, multiple aspect ratios up to 2K+, and versatile editing via circles, annotations or masks. Weights and a demo are available under a research-and-evaluation-only license.
More from Multimodal
- P2Echo: one text-prompted model segments multi-view echocardiography with high accuracy and low cost — UBC-V · 2026-11-12
- Redditor makes fast-motion short film with H3 and ships Longshot Studio v0.18.5 with face swap — R34vspec · 2026-11-12
- Reimagining Knight Rider's opening as a Netflix drama with Claude Fable 5.2 — justin_hart · 2026-11-12
- Gemini 4 Carbon builds a full 3D aquarium in Blender and Three.js in ~6 hours — sunjiao123sun_ · 2026-11-12
- Awaken Open-Sourced: Turn a Half-Duplex Speech LLM Full-Duplex With Just a Small LoRA — alexcovo_eth · 2026-11-12
- MiniMax H3 Image targets e-commerce: sketch to 4K campaign shots in ~10 seconds — MiniMax_AI · 2026-11-12