Qwen-Image model releases

last edited

This page tracks 2026 releases from the Qwen-Image line, with the Anima finetune family, SDXL checkpoints, and video tools as related context. Related coverage includes image generation breakthroughs, AI art model reference, and Kohya SS updates. Portrait checkpoints appear in the CivitAI catalog.

Qwen-Image generations

Qwen-Image-2.0 launched on February 10. It merges generation and editing into one model and accepts instructions of up to 1K tokens. The launch post describes native 2K output. These weights are closed.

Alibaba published the Qwen-Image-2.0 report on arXiv on May 11. Qwen-Image-2.1 followed on September 20 with open weights under a research license. It adds native RGBA output.

Qwen-Image-3.0 was announced on July 21. It accepts prompts near 4,500 tokens. General availability followed on August 5 through the hosted interface. No open weights have been released for it.

SDXL and Anima checkpoints

Anima, from circlestone-labs, is a finetune family. The first public preview appeared on January 31. Version 1.0 arrived on May 15. The license is non-commercial, and its model page says it does not work well at realism.

Realism Engine v1.1 was updated on February 14. This SDXL merge targets neutral photorealism. RDBT Anima v0.6d arrived on February 13 with CFG distillation. ClarityForge_XL v1.0 followed on February 11, emphasizing skin texture and stable eyes.

Video tools and ecosystem

Alibaba released Wan-Animate-2 on August 7 under a permissive license. Its diffusion-transformer architecture converts driving video into character animation. Base and distilled weights ship with inference code.

ComfyUI v0.33.1 arrived on August 13. It added partner nodes, Anima tuning blocks, and stability fixes. By June, Civitai had eight video engines, including Kling and Veo.

Further coverage appears in the ComfyUI ecosystem and platform activity pages. Commercial examples are collected under commercial use cases, and Stability AI has its own page.

Sources