A new Chinese AI model named GLM-5.2 is generating significant buzz in Silicon Valley, with tech leaders praising its coding capabilities. The open-source large language model, developed by z.AI, is designed for long-running coding tasks and agentic workflows.
The excitement recalls the industry's reaction to DeepSeek's R1 over a year ago, which established China as a major competitor in the AI race. GLM-5.2 operates on a 1 million token context window, placing it alongside Anthropic's Claude Opus 4.8 and OpenAI's GPT 5.5 according to the company.
"Genuinely impressed, almost shocked, at how good GLM-5.2 by @zai_org is at coding. This changes things," Vercel CEO Guillermo Rauch wrote on X. Across social media, investors and founders have expressed surprise at the model's speed and capability since its launch last week.
The model's performance on coding tasks has drawn particular attention for its ability to handle extended workflows. This breakthrough could intensify competition in the open-source AI space, offering developers a formidable alternative to proprietary Western models.
However, independent benchmarks verifying these claims remain limited, and the model's long-term performance against established competitors has not been thoroughly assessed. Some observers caution that early hype may outpace real-world utility.