Chinese AI startup Zhipu claims its new model outperforms Anthropic in security test
Zhipu, a Chinese AI firm also known as Z.ai, says its GLM-5.3 model beat Anthropic’s Mythos 5 on a cybersecurity benchmark and narrowed coding gaps, underscoring rapid progress in China’s open-weight AI.
The startup announced that GLM-5.3 surpassed Anthropic’s Mythos 5 in a key security test and closed much of the coding performance gap, reflecting fast iteration since its June release, one outlet noted. At the same time, the White House is drafting rules to force nations to pick a side in the U.S.-China AI rivalry, and the Trump administration has voiced concerns about China’s open-weight models while easing some chip-sale restrictions, a move described as inconsistent by a Council on Foreign Relations analyst.
Why it matters
China’s AI gains are narrowing the tech gap with the West, prompting U.S. policy responses in the AI rivalry.
In this story
