Key facts
- Z.AI has released a new large language model named GLM-5.2.
- GLM-5.2 rivals Anthropic's Claude Opus 4.8 on benchmarks.
- GLM-5.2 outperforms GPT-5.5 on benchmarks.
- GLM-5.2 was trained entirely on Huawei Ascend chips.
- The training of GLM-5.2 bypassed NVIDIA hardware.
- GLM-5.2 is available under an MIT license.
Z.AI, a Chinese artificial intelligence company, has introduced its latest large language model, GLM-5.2. This new model is positioned as a direct competitor to leading AI systems, demonstrating performance that closely rivals Anthropic's Claude Opus 4.8 on various benchmarks. Furthermore, GLM-5.2 reportedly outperforms GPT-5.5 in these evaluations. A key development in the creation of GLM-5.2 is its training process, which was conducted entirely using Huawei Ascend chips. This approach signifies a move away from the industry's prevalent reliance on NVIDIA hardware for AI model development. Z.AI has made GLM-5.2 available to the public under the terms of an MIT license, facilitating broader access and potential further development by the AI community.