In recent weeks, Silicon Valley has been hailing an AI model released by a lab in China. The program, called GLM-5.2, has been called “miracle,” “very good,” and “step by step.” Billionaire venture capitalist Marc Andreessen posted on Guillermo Rauch, the CEO of AI coding platform Vercel, said he was “really impressed, almost shocked” by GLM-5.2’s coding capabilities. Or as an AI founder told me at a recent dinner party in San Francisco: “Praise GLM-5.2.”
In a way, GLM-5.2 is China’s answer to Claude Code, Anthropic’s agent assistant that reshaped the AI boom. This has been the year of AI agents – tools that don’t just chat but promise to do things on your behalf, be it coding a website or booking a vacation. Chinese AI models have continued to improve, but none have yet proven powerful or consistent enough to be used as agents. Now GLM-5.2, developed by Chinese company Z.ai, competes with some of the top offerings from OpenAI and Anthropic and is ahead of Google Gemini in many ways. And it is many times cheaper.
Despite Silicon Valley’s awe of GLM-5.2, a low-cost competitor couldn’t have arrived at the worst time for America’s groundbreaking AI labs. After successfully convincing American companies to try their products, OpenAI, Anthropic and Google are now struggling to prove that their tools are worth the money. Using these bots can be very expensive, incurring thousands of dollars in bills per employee per month. Uber has reportedly spent its entire 2026 budget on Anthropic models in just a few months. Other major technology companies such as Meta, Amazon, Tesla and Adobe are also reportedly taking stricter action against their employees’ use of AI. According to reports, Citi blocked its employees’ access to OpenAI and Anthropic’s most expensive models at one point 404 media (which Citi has disputed).
Even as political leaders and technology executives portray the United States as being in the midst of a contentious technology race with China, Americans can still draw on Chinese models. While it’s still too early to tell whether GLM-5.2 is truly capable of replacing America’s world-class AI agents, any company or developer balking at the cost now may have an alternative. The rollout of GLM-5.2 poses a business dilemma for Silicon Valley – and potentially a national security dilemma for the entire country.
In some ways, the American AI industry has been here before. Many previous impressive Chinese AI models have not sparked a mass exodus of OpenAI and Anthropic customers, with one major exception: After DeepSeek launched a cheap AI model in January 2025 that rivaled the best in America, adoption of Chinese AI models skyrocketed. According to a study by RAND, the share of global web traffic to Chinese AI models increased from around 3 to 13 percent within two months. OpenAI, Anthropic and Google quickly responded to DeepSeek with cheaper models of their own, and then the rise of AI agents earlier this year seemed to put American labs back on top.
Even before Z.ai launched its new model, there were quiet signs of a shift to cheaper Chinese models amid growing concerns about AI bills. In the first five months of 2026, DeepSeek adoption among the 70,000 U.S. firms that use Ramp, a financial operations platform, rose from 0.1 to 0.3 percent, Ara Kharazian, Ramp’s lead economist, told me. Kharazian added that 6 percent of Ramp customers who spend money on AI use third-party platforms that provide access to many different AI products. On OpenRouter, one of these platforms, the six most popular AI models are Chinese; In less than a month, GLM-5.2 is already in fifth place. This data doesn’t capture all software developers and companies that directly download Chinese AI models, which are usually open source, and configure them on their own computers – that is, largely contentious startups and academics that don’t have the budget to use a fintech service like Ramp. According to popular AI platform Hugging Face, Chinese models accounted for almost half of all open source AI downloads from February 2025 to 2026.
In the meantime, OpenAI and Anthropic are bringing more and no less expensive products onto the market. And soon they will likely have even more competition. The hype surrounding Claude Code and the general AI agent frenzy began seven months ago, which roughly shows how far behind Chinese AI companies are in model development. China’s other labs, including DeepSeek and Moonshot AI, will almost certainly release similarly powerful and low-cost AI agents soon. Coinbase, a popular crypto company, claims to have almost halved its AI spending by ditching cheaper models like GLM-5.2 and Kimi, another popular Chinese bot. “The scenario to worry about is that China has good enough models at a quarter of the price,” Kyle Siler-Evans, an AI researcher at RAND, told me. “I think that’s probably the future we’re headed toward.”
Kharazian warned against exaggerating this narrative. While some technology-focused companies are scaling back their AI spending, most American companies are already spending very little on the technology: the average Ramp customer spends just $11 per employee on AI. “That doesn’t mean the rest of the market won’t go in that direction,” Kharazian said, but Anthropic and OpenAI will have “ample time to respond with competitive pricing” — as they did with DeepSeek.
Perhaps the biggest obstacle to GLM-5.2’s rise, at least in the US, will have little to do with model features or pricing. There are serious concerns that Chinese companies are essentially using their AI to siphon off sensitive data and steal corporate secrets. The perceived risks of using Chinese technology could have a profound chilling effect on the customers of any U.S. company — not to mention the uncertainty created by the possibility of federal regulation. Keep in mind that Chinese electric vehicles are better and cheaper than their Western counterparts in every way, but cannot be purchased in the US; One possibility is a near future in which Americans will be functionally or legally prohibited from accessing the lowest-cost AI models.
The biggest impact of GLM-5.2 and the impending onslaught of cheap Chinese AI agents could also be geopolitical rather than economic. Whatever soft power the US gains from its technology may be waning as software developers in countries with less tense relations with China turn to DeepSeek and GLM-5.2 in greater numbers (think of how the EU, UK and Canada import Chinese electric vehicles). Although Big Tech is spending increasingly large sums on advanced AI chips to train their models – chips that are banned from export to China – the gap between US and Chinese models has not widened; If anything, it may shrink. That means any military, economic, cyberoffensive or other advantage that AI can provide the nation could disappear. For the first time since DeepSeek, America’s lead in AI competition is in real danger of faltering.
https://www.theatlantic.com/technology/2026/07/glm-5-2-china-cheap-ai-agents/687828/
