Alibaba launches Qwen-Image-3.0 with 4.5k-token prompts and 10px text rendering
千问大模型 · wechat · 2026-07-21
Alibaba’s Qwen team officially launched Qwen-Image-3.0, a new image generation foundation model built around three core claims: richer content, finer detail, and broader world knowledge.
What it can do
- Richer content: accepts prompts up to 4.5k tokens, enabling complex layouts such as newspapers, storyboards, and exam papers.
- Finer detail: renders 10px text clearly and reproduces pores, hair, and other micro details with high fidelity.
- Broader knowledge: supports native rendering in 12 languages, realistic web/game/live-stream UI mockups, and more world knowledge.
Notable demos
- A single prompt generating a dense 9-panel infographic with math, science, and text-heavy content.
- Nested UI scenes, such as VS Code → Qwen chat → social chat → coffee poster in one image.
- Academic paper pages with readable formulas and notation.
- Handwritten annotations on book pages.
- Restoration of damaged traditional paintings while preserving brushwork and style.
The model is now available via invite-only API access on Alibaba Cloud Bailian and the Qwen platform, with QwenStudio and the Qwen app coming soon for free trials.
More from Multimodal
- 10 practical ways to run Astra computer use on Premiere Pro for video editing — beechinour · 2026-10-27
- MiniMax H3 runs fully on a Galaxy S26: 5-second video with audio in ~7 minutes — BidFunny4862 · 2026-10-27
- Krea ships Premiere Pro plugin to generate video and SFX inside the timeline — lmoroney · 2026-10-27
- Wan2GP lands on Pinokio: 1-click AI video on 6GB GPUs — cocktailpeanut · 2026-10-27
- WanGP v18.81 adds Tencent's Hy-Nomas world model, playable on 8GB VRAM — cocktailpeanut · 2026-10-27
- Butter.video wins over skeptical creator with hundreds of tasteful timeline 'blocks' for video effects — beechinour · 2026-10-27