Chinese AI models are gaining on Anthropic and OpenAI

Two weeks ago, artificial intelligence company Antropic shut down its two most powerful AI systems after an unexpected request from the US government to restrict access to them.

A few days later, a Chinese startup, Z.ai, released an AI model that is nearly as powerful as the company’s Anthropic, Fable, and Mythos models. But Z.ai’s new technology costs much less to use, and no one in the United States has placed restrictions on it. He quickly landed on the closely watched world rankings 10 most popular models.

Z.ai is at the forefront of a wave of powerful but low-cost artificial intelligence from China that is challenging OpenAI, Anthropic and Google’s lock on the industry. The six models now on the AI ​​chart were developed in China.

Z.ai’s new model, the GLM-5.2, arrives just as American businesses realize they need to find ways to reduce AI spending. He also landed as Silicon Valley executives grew concerned that the Trump administration was leaning toward regulating the technology.

“With the limited Fable, the gap between the US and China is very small,” said Rehaan Ahmad, co-founder of Silicon Valley startup alphaXiv, who has been using the new Z.ai model for more than a week.

The Chinese models still face two major obstacles to widespread use in the United States: concerns about their ties to the Chinese government and complaints that Chinese companies have unfairly used American technology to make these cheaper models. But their low cost is getting conversions.

About 18 months ago, Chinese startup DeepSeek shocked Silicon Valley by showing that it could create effective artificial intelligence much more cheaply than many of its American counterparts. Z.ai does something similar. For certain tasks, GLM-5.2 costs about an eighth as much as Anthropic’s Claude Opus 4.8, which came out shortly before Fable and Mythos, according to OpenRouter, an AI ranking start-up.

Like most of the top Chinese models, the GLM-5.2 is open source software, which means that anyone can use and modify it for free. This makes it much cheaper to use, although not as strong as what American companies have created.

“Do you need to drive a Ferrari everywhere?” asked Vivek Ramaswami, an early stage investor at Madrona Venture Group. “Probably not.”

Z.ai did not respond to requests for comment.

GLM-5.2 is particularly good at generating computer code and powering AI agents, digital assistants that can use other software to perform tasks. Z.ai’s technology is now the third most widely used in the world for AI tasks, said Anastasios Angelopoulos, chief executive of ArenaAI, which tracks millions of AI users.

The biggest cloud computing providers, including Microsoft and Amazon, already offer access to some systems from Z.ai, DeepSeek, MiniMax and other Chinese start-ups. Microsoft has also considered adding DeepSeek’s latest model as an option to power one of its own products, which now runs on technology from Anthropic and OpenAI, said two people familiar with the considerations, who spoke on condition of anonymity because they were not authorized to discuss them publicly.

There were talks reported previously from Axios.

Microsoft, Anthropic and OpenAI declined to comment.

Some software developers are reluctant to use the AI ​​system Z.ai offers from computers in China because they fear sharing data with the company or the Chinese government. They are also wary of China’s efforts to censor its artificial intelligence systems or run afoul of US export restrictions.

Z.ai was added to the Ministry of Commerce’s trade blacklist in 2025. Company records show that several of the company’s shareholders are controlled by a Chinese government agency that oversees the country’s defense industry.

Companies can still use this model without sending data back to China in violation of U.S. export rules if they are careful about how they set up their systems, said Wei Chen, chief legal officer at Infoblox, a network security company.

“The Chinese models don’t have the same limitations if you host them yourself or go through another provider,” alphaXiv’s Mr. Ahmad said. “Right now there are additional restrictions on models from Anthropic.

After DeepSeek was released in 2025, governments around the world passed regulations restricting its use due to data security concerns. But so far, GLM-5.2 has not raised a similar alarm, Ms. Chen said.

Anthropic and OpenAI have accused Chinese companies of improperly extracting data from their AI systems to speed up the development of Chinese technology. On Wednesday, Antropic sent a letter, seen by The New York Times, to Senators Tim Scott, R-South Carolina, and Elizabeth Warren, D-Massachusetts, accusing Chinese tech giant Alibaba of a “brazen” and “illegal” attempt to copy its technology through 24,000 fraudulent accounts.

Alibaba declined to comment.

When Mustafa Suleyman, head of Microsoft’s AI lab, unveiled a set of new models this month, he emphasized that they were built from scratch on data that the company licensed commercially.

“That means you can put it into production in a very credible way with complete confidence,” he said.

(The Times sued OpenAI and its partner, Microsoft, accusing them of copyright infringement of news content related to AI systems. They denied the claims.)

Using data from one system to train another—a process called distillation—is common in AI development. But Anthropic and OpenAI’s terms of service prohibit anyone from secretly collecting data for distillation. It is unclear whether Z.ai used distillation to develop its technology.

But distillation alone cannot build a cutting-edge AI system. It also requires several other complex techniques, said Charles O’Neill, head of model training at Baseten, which sells access to GLM-5.2.

“This story that all the capabilities of these models come from Anthropic is not as true as people say,” Mr O’Neill said.

Chinese AI startups can offer their models as open-source technology at much lower prices, in part because the industry has benefited from years of support from the Chinese government, which sees AI as a critical engine of economic growth.

Many executives said that US companies should not open source their technology because it could be used in harmful situations. But other experts say that if regulators stifle open-source technology in the United States, China will gain a significant head start.

Because China produces most of the most powerful open-source systems, they say American developers will build their software on those technologies. In the long term, this could put China at the center of AI development.

Some argue that Chinese systems will always lag behind the best US models because US export controls restrict the flow of specialized computer chips needed to train AI technologies. Z.ai and other Chinese startups are spending millions to access chips in data centers outside of China.

Z.ai’s Hong Kong filing shows that in the first half of 2025, the company spent more than seven times its revenue on expenses that essentially boiled down to fees for such computing services.

Still, experts estimate that China lags American companies by only six months or less.

“There was speculation that export controls would eventually bite and the gap between US border models and their Chinese counterparts would widen, but the GLM is pushing things in the opposite direction,” said Jeffrey Ding, an assistant professor at George Washington University who specializes in new technologies and international relations.

And with Fable and Mythos out of the way, many businesses realized the importance of having alternatives.

“There’s a bit of a fear of loyalty in large organizations,” said Justin Summerville, who runs data analytics at OpenRouter. “Who knows what the top model will be like in three weeks?”