China’s Z.AI Releases GLM-5.2: A Model That Rivals Claude Opus—Using Zero Nvidia Chips
Z.ai released GLM-5.2 on June 16, positioning it as a major upgrade over GLM-5.1 and a strong open-source rival to top closed models. The model posted strong results on technical benchmarks, nearly matching Claude Opus 4.8 on FrontierSWE and outperforming GPT-5.5 on SWE-bench Pro. It also became the highest-ranked open model in the Artificial Analysis Intelligence Index. A key angle is hardware and cost: GLM-5.2 was trained entirely on Huawei Ascend chips, with no Nvidia hardware, and reported training costs were estimated at about $25 million. It is a 744B-parameter mixture-of-experts model with a 1 million-token context window, far beyond GLM-5.1’s 200K. Open weights are available under the MIT license. For developers, the huge context window makes whole-repo work, long code refactors, and agentic pipelines much easier. Pricing is also far lower than Claude’s. Local deployment is possible via quantized versions, though it still requires very large memory.
