NVIDIA launches Jetson Thor robotics computers for physical AI systems
Investing.com -- Alibaba (NYSE:BABA) has released Qwen-Image, a 20B MMDiT image foundation model that delivers significant advances in complex text rendering and precise image editing capabilities.
The new model, which users can access through Qwen Chat by selecting "Image Generation," features superior text rendering abilities that handle multi-line layouts, paragraph-level semantics, and fine-grained details. It supports both alphabetic languages like English and logographic languages such as Chinese with high fidelity.
Qwen-Image also offers consistent image editing through an enhanced multi-task training paradigm, achieving exceptional performance in preserving both semantic meaning and visual realism during editing operations.
According to Alibaba, the model outperforms existing solutions across multiple public benchmarks for both generation and editing tasks, including GenEval, DPG, OneIG-Bench, GEdit, ImgEdit, and GSO. It particularly excels in text rendering benchmarks such as LongText-Bench, ChineseWord, and TextCraft, where it significantly outperforms current state-of-the-art models.
The company demonstrated Qwen-Image’s capabilities through various examples, showcasing its ability to render complex text in different scenarios. These include accurately generating Chinese characters on shop signs with proper depth of field, creating detailed English text on book covers and information slides, and handling bilingual content with ease.
Beyond text processing, Qwen-Image supports a wide range of artistic styles from photorealistic scenes to impressionistic paintings, and offers various editing operations including style transfer, additions, deletions, detail enhancement, text editing, and character pose adjustment.
Alibaba stated that Qwen-Image aims to promote the development of image generation, lower technical barriers to visual content creation, and inspire innovative applications. The company is inviting community participation and feedback to build "an open, transparent, and sustainable generative AI ecosystem."
The model is scheduled for launch in August 2025.
This article was generated with the support of AI and reviewed by an editor. For more information see our T&C.