Chinese startup Z.ai has launched the GLM 4.5 family of AI models, boasting superior performance over Anthropic's Claude 4 and OpenAI's o3 across multiple benchmarks. The GLM-4.5, with 355 billion parameters, and its Air variant are said to unify capabilities in coding, reasoning, and agentic tasks. Early user feedback indicates exceptional coding performance. “GLM-4.5 is absolutely crushing it for coding,” noted one user, clarifying Z.ai’s challenge to Western AI leaders.