June 25, 2026, (Inside AI) — Chinese AI startup Z.ai has stunned the global tech community by releasing its GLM-5.2 model, which nearly matches the performance of top U.S. closed-source systems from Anthropic and OpenAI on public benchmarks. The launch came just one day after Anthropic disabled worldwide access to its most advanced models, intensifying scrutiny on AI sovereignty.
The Beijing-based company, also known as Zhipu AI, now plans a dual listing in Shanghai to fund its pursuit of artificial general intelligence. Its shares have rallied over 2,000% since a blockbuster Hong Kong debut in January, pushing market capitalization past HK$1 trillion ($128 billion) this week.
GLM-5.2 holds fourth place on Artificial Analysis' LLM intelligence leaderboard and second spot on Code Arena's front-end coding leaderboard. It operates at roughly one-sixth the cost of closed U.S. frontier models, according to company data.
"This model is comparable to the top closed models," said Qinkai Zheng, technical lead of the CodeGeeX team, at the firm's Beijing headquarters.
"It's the first time that an open-source model really delivers very solid coding and agent performance that can compare with the leading proprietary AI companies like Anthropic and OpenAI."
The sudden withdrawal of closed U.S. models has sparked anxiety among allies. At a G7 summit last week, leaders from Canada and France criticized over-reliance on U.S.-controlled AI infrastructure. Z.ai's breakthrough marks the first time a Chinese open-source model has narrowed the frontier gap so dramatically, having already surpassed U.S. open-source models like Google's Gemma and Meta's Llama series despite chip constraints.
Analysts previously estimated Chinese AI models lagged four to six months behind U.S. leaders. GLM-5.2's performance suggests that gap may be closing faster than expected.
The model specializes in coding and complex long-horizon tasks, with 750 billion parameters and a 1-million token context window. Since February, Z.ai has adapted its GLM-5 series to run on domestic semiconductors, including Huawei Ascend clusters, after U.S. export controls tightened access to advanced Nvidia chips.
"We are trying our best to improve our infrastructure and to make the model more efficient on different kinds of chips," Zheng said, declining to disclose whether training used domestic or foreign chips.
Despite fierce price competition in China, Z.ai has raised prices for its frontier models multiple times this year, reflecting strong enterprise and public sector adoption. JP Morgan projects revenue will balloon this year, with profitability expected in 2028. Yet stock exchange filings show Z.ai earns only a fraction of its U.S. counterparts' revenue.
"We are trying to lower the cost, but because the demand is too large, maybe in the future we will still need to increase the price," Zheng said. "But we want the model accessible to everyone."
Co-founder Tang Jie emphasized Z.ai's open-source commitment on X, calling Anthropic's abrupt model pull "deeply regrettable."
Zheng said future models will target long-horizon tasks and self-evolving autonomous agent systems. The next iteration, GLM-5.5, is expected in August and could become a key milestone in Chinese frontier AI.
