Alibaba’s Qwen team has launched Qwen-Image-2512, a fully open source, enterprise-grade AI image model designed to rival Google’s Nano Banana Pro.
Alibaba’s Qwen AI research team has released Qwen-Image-2512, a fully open-source, enterprise-grade AI image generation model positioned directly against Google’s Nano Banana Pro, also known as Gemini 3 Pro Image.
Licensed under Apache 2.0, Qwen-Image-2512 allows free commercial use, modification, fine-tuning, and self-hosted deployment, offering enterprises an open alternative to proprietary image models from Google and OpenAI. The release addresses core enterprise requirements around predictable costs, deployment sovereignty, data governance, and localisation—areas where closed systems remain tightly coupled to vendor infrastructure.
The model is designed for text-heavy, structured visual generation, including infographics, slides, menus, posters, and multilingual enterprise documentation. It supports both Chinese and English prompts, with improved text accuracy and layout consistency.
The December 2512 update introduces key performance gains across human realism, natural texture fidelity, and structured text rendering. Facial features, posture, and environments display reduced ‘AI look,’ while landscapes, materials, and surfaces show finer detail. Embedded text and mixed text-image layouts are rendered with greater legibility and instruction adherence, enabling production use across ecommerce, education, and visualisation workflows.
In blind, human-evaluated testing on Alibaba’s AI Arena, Qwen-Image-2512 ranked as the strongest open source image model and remained competitive with leading closed systems, reinforcing its positioning as production-ready.
The model is accessible via Qwen Chat, hosted demos on Hugging Face and ModelScope, and through Alibaba Cloud Model Studio as qwen-image-max, priced at $0.075 per image. The hybrid open-core and managed API approach reflects real-world enterprise deployment patterns.
