Alibaba’s Qwen team released a new image generation artificial intelligence (AI) model last week. Dubbed Qwen VLo, it is a successor to the Qwen 2.5 vision language model and comes with several upgrades compared to the older models. The latest AI image model supports both text-to-image and image-to-image generation. It also supports text input in multiple languages, including English and Chinese. Apart from image generation, the AI model is also capable of making inline edits to generated images as well as input images.
Qwen VLo Accepts Prompts in Multiple Languages
In a post on X (formerly known as Twitter), the official handle of the Qwen team announced the release of the new model. The model’s technical name is Qwen3-235B-A22B, and it is available on the company’s chat interface for free here. Users can also use the model without logging in.
Gadgets 360 staff members tested out the AI model and found its image generation capability to be on par with Google’s Imagen 2. The instruction following and image output quality is slightly lower than Imagen-3 and OpenAI’s GPT-4o-powered image generation feature. However, its generation time is faster than both of them, and it has a higher rate limit than them.
On its GitHub page, the company said that the Qwen VLo comes with improved image understanding, which enables it to make better inline edits without distorting the structural integrity of the input image. This also improves the overall quality of the output. The model also better understands vague and open-ended prompts, and can generate images that are aligned with user expectations.
Apart from image generation and editing, the Qwen VLo can also perform image annotation-related tasks such as edge detection, segmentation, prediction mapping, and more. The company said the future version of the model will also be able to accept multiple input images and combine them based on user requests.
Text rendering has also been improved with the latest AI image generator. We were able to generate accurate text across different fonts in our testing of the model. Finally, the Qwen VLo also supports images with dynamic aspect ratios as input, including extreme ratios such as 4:1 and 1:3. The company plans to add the feature to generate images in different aspect ratios soon.