X Square Robot open-sourced Wall-OSS-0.5, a Vision-Language-Action model for embodied intelligence. It achieved zero-shot deployment on real robots across 17 challenging tasks without fine-tuning, scoring up to 82 on unseen flexible object tasks. Compared to industry benchmark π0.5, it led by 17.5 points on average and showed near order-of-magnitude improvement in precision operations. Key innovations include gradient bridging, visual alignment tokenizer, action space supervision, and DMuon distributed optimization. The model weights, training code, and dataset interfaces are fully open-sourced.