GLM 5.2 Impresses Tech CEOs: "This Changes Things

A new Chinese AI model, GLM 5.2, is making a significant impression in the Western tech community. Released by Z.AI, this open-source model has been called "genuinely impressed" by Vercel CEO Guillermo Rauch, who said, "This changes things." Box CEO Aaron Levie notes the "remarkable" progress of open weights AI, seeing it as a positive for the entire applied AI sector. Jeremy Howard of Answer.AI and Fast.AI calls GLM 5.2 "a marvel," comparing it to top models like Claude Opus 4.8 and GPT 5.5 for its nuance and reliable long-context handling. Mat Velloso, formerly of Meta and Google DeepMind, even calls it the first open model that can be a "daily driver." GLM 5.2 scores 62.1 on SWE-bench Pro, surpassing GPT 5.5, and leads open-weights models by a seven-point margin on the Artificial Analysis Intelligence Index. These strong benchmark results are being confirmed by experts using the model for practical tasks. The model's improvements come from training advancements, not just scale, with its context window expanding to one million tokens. This signals a narrowing gap between open and closed AI models, potentially accelerating innovation across the industry.