Alibaba Releases Qwen-Image-3.0 With Improved Text Rendering
Early posts on X point to sharper text layout and unusually long prompts, though the bigger capability question is still mostly being argued from examples.
Entities: Alibaba, Qwen-Image-3.0
Alibaba's new Qwen-Image-3.0 is getting early attention on X after Lentils80 posted example outputs, including an academic-paper-style image they said showed striking text rendering, layout and improved world knowledge. A quote-post from teortaxesTex focused on another claim: that Qwen-Image-3.0 accepts instructions up to 4.5k tokens, which could help explain why dense, text-heavy layouts appear more coherent in the shared examples.
Combined views
11.7K
3 posts, first seen 7h ago
Alibaba Releases Qwen-Image-3.0 With Improved Text Rendering
Early posts on X point to sharper text layout and unusually long prompts, though the bigger capability question is still mostly being argued from examples.
Entities: Alibaba, Qwen-Image-3.0
Alibaba's new Qwen-Image-3.0 is getting early attention on X after Lentils80 posted example outputs, including an academic-paper-style image they said showed striking text rendering, layout and improved world knowledge. A focused on another claim: that Qwen-Image-3.0 accepts instructions up to 4.5k tokens, which could help explain why dense, text-heavy layouts appear more coherent in the shared examples.
Reactions from ranked influencers
3 posts
🚨 Alibaba released Qwen-Image-3.0 today I attached a few examples, images 2nd through 4th are generated by it The text rendering and layout on that academic paper image is very very impressive, and it seems to have improved world knowledge
I see zero progress (or change) in "generate image" in free qwen studio though not deployed I guess
Crazy stuff «And yet this remains effortless for Qwen-Image-3.0 — because Qwen-Image-3.0 raises the acceptable instruction length to 4.5k tokens» The question is whether it's a gimmick or if it can coherently reason over such complex layouts… gotta test
🚨 Alibaba released Qwen-Image-3.0 today I attached a few examples, images 2nd through 4th are generated by it The text rendering and layout on that academic paper image is very very impressive, and it seems to have improved world knowledge