Qwen-Image-2.1 open weights: 7B image model runs on RTX 3090, 2K output in ~21s
量子位 · wechat · 2026-09-21
Qwen released Qwen-Image-2.1 with open 7B weights, packing text-to-image, image editing, transparent-image generation and native 2K output into one model that can reference up to 10 images per generation. It quickly hit #1 on Hugging Face's text-to-image trending chart.
- Community benchmark on dual RTX 4090 (48GB): 21 seconds per HD image; a 2752×1536 true-2K wallpaper takes just over two minutes.
- Official comparisons show it beating Nano Banana 2.0 and GPT Image 1.5; users report fine-grained detail and smooth use as first frames for image-to-video.
- Direct transparent-background output removes the manual cutout step for design workflows.
- Caveat: released under Qwen Research License, not Apache 2.0 — public weights are for non-commercial research/evaluation only; commercial use requires separate licensing.
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