Alibaba’s Qwen team released Qwen Image 3.0 on July 21. The model accepts up to 4,500 tokens of instructions, 4.5 times the previous generation, enabling single-pass generation of complex layouts such as nine-panel infographic grids. It can render text as small as 10px and accurately reproduce LaTeX mathematical equations. The launch came without published benchmarks, open model weights, or a technical report, only with company-selected example images. The tool is accessible via chat.qwen.ai for API trials, but pricing has not been disclosed.
Source: Read the original article
