Chinese AI developers are launching models that rival the capabilities of leading U.S. labs like Anthropic and OpenAI, even as U.S. government restrictions slow American innovation. Zhipu's GLM 5.2, released earlier this month, can perform on par with top U.S. labs on some cybersecurity benchmarks, according to researchers. The timing is striking: while Anthropic faced a two-week shutdown due to export control directives and OpenAI limited its GPT 5.6 rollout following a government request, Chinese companies are reaching U.S. users with increasingly capable open-weight models.

What Are Open-Weight AI Models and Why Do They Matter?

Open-weight models are artificial intelligence systems whose underlying code and parameters are publicly available, allowing companies and developers to download them and run them on their own servers without relying on a third-party cloud service. This approach has several advantages: it reduces costs, increases privacy, and makes it easier for organizations to customize the models for their specific needs. Chinese developers have embraced this strategy, and it's proving effective at reaching U.S. customers.

The shift toward open-weight models is happening at a critical moment in the AI industry. Corporate America is moving away from what insiders call "tokenmaxxing," or allowing developers to spend on AI without restraint, toward a focus on efficiency and return on investment. This transition plays directly into China's hands, since open-weight models are typically cheaper to run than proprietary alternatives.

Which Chinese Models Are Gaining Traction in the U.S.?

Several Chinese AI models are now being adopted by U.S. companies and startups. Flo Crivello, CEO of AI startup Lindy, switched his company entirely off Anthropic's Claude models and moved 100 percent of its traffic to DeepSeek, a Chinese company that makes cheaper, open-weight alternatives. "We did it, and you could see that cost curve go down, like, crash to the ground," Crivello told CNBC.

Major companies are also experimenting with Chinese models. Shopify and Airbnb have touted the benefits of Alibaba's Qwen 3 for scaling AI features. Coinbase CEO Brian Armstrong wrote on social media last week that his company is utilizing open-weight models such as GLM 5.2 and Kimi 2.7, allowing it to cut nearly half its AI spending despite greater token use.

Travis Lanham, co-founder of AI security startup Armadin, said the models are showing improved capabilities for cybersecurity use cases like analyzing reconnaissance data and creating customer exploit code. Lanham's team is experimenting with both GLM 5.2 and Kimi K2.7 from Moonshoot AI.

How Are Chinese Models Performing Against U.S. Competitors?

According to industry analysis, GLM 5.2 is narrowing the gap with U.S. frontier labs at a remarkable pace. Venture capitalist Marc Andreessen wrote on social media that GLM 5.2 is "the first Chinese AI model to match and often beat the American big lab public AI models with no compromises." He added, "Incredible timing given current events".

Sam Bresnick, a research fellow at Georgetown's Center for Security and Emerging Technology, called the recent developments "a pretty good wake-up call." Jefferies strategist Christopher Wood wrote in a report to clients, citing industry sources, that GLM 5.2 "is almost equal to Anthropic as a competitor for the corporate market and is just one quarter of the cost in terms of cost per token".

"GLM-5.2 is almost equal to Anthropic as a competitor for the corporate market and is just one quarter of the cost in terms of cost per token," noted Christopher Wood, strategist at Jefferies.

Christopher Wood, Strategist at Jefferies

Elon Musk predicted that GLM 5.2 would reach the capabilities of Anthropic's Fable model by the first quarter of next year. Zhipu founder Jie Tang, responding to Musk, wrote, "won't take that long".

Elon Musk

Steps to Understand the Competitive Landscape in Chinese AI

• Model Capabilities: GLM 5.2 from Zhipu is matching U.S. frontier labs on cybersecurity benchmarks and is expected to reach higher capability levels within months, according to industry leaders.
• Cost Advantage: Chinese open-weight models cost approximately one quarter the price per token compared to Anthropic's Claude, making them attractive to cost-conscious enterprises.
• Accessibility: Open-weight models can be downloaded and run on company servers without relying on third-party cloud services, reducing dependency on U.S.-based AI providers.
• Real-World Adoption: Major companies like Coinbase, Shopify, and Airbnb are already integrating Chinese models into their operations to reduce AI spending.

What Are the National Security Implications?

The rise of capable Chinese AI models raises serious questions about cybersecurity and national defense. Some open-weight models can already automate many stages of a cyberattack, and industry experts worry they are only months away from running an entire operation autonomously.

Hed Kovetz, CEO of industry startup Silverfort, expressed concern about the timeline. "If the U.S. government does not let the industry take advantage of this opportunity to get ready, then when the Chinese models reach a similar level, no one will be prepared," he said.

Hed Kovetz, CEO of industry startup Silverfort

"If the U.S. government does not let the industry take advantage of this opportunity to get ready, then when the Chinese models reach a similar level, no one will be prepared," warned Hed Kovetz.

Hed Kovetz, CEO at Silverfort

The U.S. government has historically used export controls on AI chips from Nvidia and Advanced Micro Devices to keep cutting-edge AI innovation out of China's hands. However, the recent restrictions on U.S. AI companies may inadvertently give Chinese competitors more time to close the gap. Last year, the U.S. cleared Nvidia's H200 chip for export to the China region, though Nvidia said it had yet to generate revenue from those sales.

Former Trump crypto and AI czar David Sacks highlighted the strategic concern, writing that "a year ago, President Trump declared that America was in a global AI race and that the way to win it was to be pro-innovation, pro-infrastructure, pro-energy, and pro-export. President Trump was exactly right; we deviate from that strategy at our peril".

The situation underscores a fundamental tension in U.S. AI policy: restricting American companies to protect national security may inadvertently slow domestic innovation while Chinese competitors accelerate their progress. As Chinese models become more capable and cheaper to operate, U.S. policymakers face mounting pressure to balance security concerns with the need to maintain technological leadership.