Qwen VLo: Unified multimodal understanding and generation model from Alibaba

Alibaba introduces Qwen VLo, a new unified multimodal model that can both understand and generate images. The model is an upgrade from previous QwenVL series, bridging perception and creation. It uses a progressive generation method, constructing images gradually from left to right and top to bottom. Qwen VLo supports precise content understanding, open-ended instruction-based editing (style transfer, object modification, background change), multilingual instructions (Chinese and English), and dynamic resolution with arbitrary aspect ratios. It can also perform visual perception tasks like depth map prediction, segmentation, and edge detection via editing commands. The model is in preview stage and available through Qwen Chat. Limitations include occasional inaccuracies and inconsistencies.