Chinese companies are quietly turning into serious players in AI coding tools, with Zhipu’s GLM 4.7 and MiniMax’s M2.1 as two of the most interesting new models. GLM 4.7 is positioned as Zhipu’s flagship model with strong upgrades in programming and multi step reasoning, and on public coding benchmarks it lands close to leading systems like Claude Sonnet 4.5, even setting open source records on tests such as SWE bench Verified and LiveCodeBench.
MiniMax M2.1 goes in a slightly different direction, focusing not just on code correctness but also on handling complex “agent style” workflows, tool use and multi language programming, all while trying to keep responses shorter, faster and cheaper in tokens. In everyday use, China’s AI landscape looks different from the US: ByteDance’s Doubao dominates as a mass market assistant with well over 100 million monthly users, ahead of Tencent’s Yuanbao and DeepSeek, and Ant Group pushes its A Fu health app and Alibaba’s Qwen deeper into consumer services.
Overall, Chinese coding models are now competitive with strong Western mid to high tier models on many benchmarks, but leadership at the very top of the charts still mostly belongs to US systems.
