Alibaba Unveils Image Generation AI with Enhanced Practicality

Alibaba has unveiled 'Qwen-Image 3.0,' a next-generation image generation artificial intelligence (AI) model ready for immediate deployment in actual work environments. If Qwen-Image 1.0 pursued 'precision' and 2.0 focused on 'completeness and diversity,' this 3.0 version completes three key pillars: 'rich content,' 'authentic details,' and 'deep knowledge.'

This model, which generates desired images from text prompts as if taking a photograph, supports 12 languages, including Korean, and has significantly increased text length limits. It features enhanced performance designed for immediate use in practical tasks such as education, content creation, and e-commerce operations. It can also render over 100 different artistic styles.

A major update is the expansion of the prompt processing length to a maximum of 4,500 tokens. Consequently, complex academic papers, newspapers, and infographic layouts containing intricate formulas and charts can be rendered in a single generation process.

The model clearly recognizes text as small as 10 pixels (px) and depicts details like skin pores with ultra-high precision. It also seamlessly reproduces diverse web and game user interfaces (UI), such as actual webpages or game screens.