ModelScope Civision adds Qwen-Image-2.1 inference and LoRA training
ModelScope says it now supports both inference and LoRA training for Qwen-Image-2.1 in Civision.
The announcement links to modelscope.ai/civision.
Related AI news
inclusionAI open-sources two 6B Ming-Image-0.1-Design models
inclusionAI has open-sourced the Ming-Image-0.1-Design family under the MIT License: two complementary 6B models, Design and Layer, for visual-design workflows. ModelScope says Design ranks first among open-weight models on the Artificial Analysis UI/UX Design leaderboard, and that Layer leads all 12 evaluated Crello settings while running 4.3× faster than the evaluated 20B open-weight baseline.
Qwen introduces RecreationBench for hybrid computer-use agents
ModelScope said Qwen has introduced RecreationBench, a 250-task benchmark for Hybrid Computer-Use Agents across Ubuntu, macOS, Windows, Android, and Web. Agents must explore a running app, recreate it in code, and pass programmatic tests plus VLM-based visual evaluation.
Shanghai AI Lab and SJTU's LUMIA Lab release 8.9B open-weight model NCP-ArchPreview
NCP-ArchPreview is released under Apache 2.0. The labs report that, after training on 5.73 trillion Dolma 3 tokens, it matches OLMo-3-7B’s final Stage 1 loss with 51.3% of the tokens, a 1.95× convergence gain.
Frame-Lock LoRA for Qwen Image 2.1 is meant to keep edits on the original frame
Wildminder says the English-trained LoRA produces zero movement for the same prompt and edit, blocking height, sideways, and seed drift as well as unwanted repaints. Comic and anime restyles may still shift by up to 12.9 pixels. The model card is Qwen-Image-2.1-Consistency-LoRA on Hugging Face.
Qwen Image 3 Pro leads the first Memebench bracket with 35%
In the first Memebench bracket, Qwen Image 3 Pro took 35% of the vote, ahead of Nano Banana Pro at 33%, Riverflow V2 Pro at 23%, and GPT Image 2 at 9%.