Home›Crypto›Regulation›Alibaba releases Qwen Image 3.0 for dense, single-pass…

Alibaba releases Qwen Image 3.0 for dense, single-pass infographic generation

The model can take up to 4,500 tokens of instructions in one request and render text down to 10 pixels, according to Alibaba.

Reported · Written up by our desk · Regulation

Alibaba’s Qwen team released Qwen Image 3.0 on July 21, positioning the image model as a “useful” productivity tool rather than a system focused only on visually appealing outputs.

According to Decrypt, the model accepts up to 4,500 tokens of instructions, enabling single-pass creation of complex layouts such as newspapers, storyboards, and dense infographic grids without stitching multiple images together.

The announcement also highlights “authentic details,” including the ability to render text as small as 10 pixels and accurately handle LaTeX across full academic-paper mockups.

While the release went live at chat.qwen.ai, Decrypt reports it shipped without open model weights, benchmarks, or a technical report, and API pricing was not disclosed at launch.

More like this

Sources