Z.ai’s open-weights GLM-5.2 beats GPT-5.5 on multiple long-horizon coding benchmarks for 1/6th the cost

Z.ai launched GLM-5.2, a 753-billion-parameter open-weights LLM aimed at long-horizon autonomous coding and engineering tasks. The model is available through Hugging Face, Z.ai’s API, and coding tools, with claims of strong benchmark performance, a stable 1-million-token context window, and low-cost enterprise plans.

GLM-5.2 reflects the growing push toward powerful models that enterprises can host, customize, and govern outside closed vendor ecosystems. Its MIT license and local deployment options may appeal to companies worried about regulatory risk, export controls, data security, and rising proprietary AI costs.