Qwen Image 2.1 beats Google Nano Banana 2.0 with minuscule 7B parameter model (www.tomshardware.com)

Alibaba Cloud has launched Qwen Image 2.1, a lightweight image generation AI model with a mere 7 billion parameters, enabling it to run efficiently on older consumer graphics cards like the RTX 3090. This new model reportedly outperforms several closed-weight alternatives, including Google's Nano Banana 2.0, in internal benchmarks. Key features include native transparency support and the ability to generate unified images from multiple reference images, enhancing its utility for artists and print-on-demand applications. However, the licensing agreement has been revised to prohibit commercial resale without a separate license, raising concerns about its open-weight status. Early adopters have reported impressive performance, generating quality images on various GPUs, including older models. While it excels in editing versatility, particularly with multiple reference images, some users have noted reduced speed in this functionality. While further testing is needed to validate the developers' claims, Qwen Image 2.1’s compact size and capabilities position it as a formidable competitor against larger models. Its licensing changes may affect community perception, but for most users, the ability to run powerful image generation locally without cloud reliance remains a significant advantage.