GLM-5.2: Open Weights, Near-Frontier Intelligence — Zixuan Li, Z.ai

The post introduces GLM-5.2, an open-weight model that rivals proprietary frontier models like Claude Opus on long-horizon coding and agentic benchmarks. It details the rationale behind releasing near-frontier models as open weights, emphasizing benefits for enterprise deployment, custom fine-tuning for specific domains (law, finance, security), and collaborative development with the community. Additionally, it announces Z Code, a new coding harness built for GLM-5.2 and compatible with other frontier models, aiming to accelerate AI development workflows.