August 19, 2026 GLM-5.3: 50% Coding Boost via Post-Training Scaling Delayed Over Cyber-Offensive Risks AIChinese LLMs
