Curious how Qwen Image 2.1 performs on less common datasets compared to SDXL. Always good to see more open models.