HONG KONG (AP) — Chinese There is a new hot product in the United States: artificial intelligence.
San Francisco-based Raffi Krikorian, chief technology officer at Mozilla, which runs the Firefox browser, has moved on to the Chinese AI startup Moonshot’s Kimi K3 for most of their daily activities in the days following the launch of the powerful new model more than a week ago.
“It seems faster,” he said of the K3, comparing it to the acclaimed and more expensive Claude Fable chatbot from San Francisco company Anthropic. He previously used another high-performance Chinese model, Z.ai’s GLM-5.2, for routine tasks like managing his calendar, documents, and email.
Krikorian is one of a growing number of Americans embracing Chinese artificial intelligence systems that are gaining ground around the world because they are more affordable and increasingly efficient. US companies such as cryptocurrency exchange Coinbase have said they will switch to Chinese AI models to help reduce costs.
Their popularity has disappointed some US tech giants, but unless they are banned outright they will continue to attract interest from independent software developers in the US and elsewhere.
US-led restrictions More sanctions could be coming to block China’s access to some of the world’s most advanced technologies, including cutting-edge artificial intelligence chips, and to protect American intellectual property, US Treasury Secretary Scott Bessent has warned.
US President Donald Trump’s administration on Wednesday accused Moonshot of using “secret” but not necessarily illegal methods to build K3, based on Anthropic’s Fable. Some US politicians and artificial intelligence companies, including Anthropic, have accused Chinese startups of illegal activities. “distillation” It’s a claim that Beijing rejects as “baseless”.
China’s AI models are cheaper and ‘good enough’
The meteoric rise of cutting-edge Chinese AI models this year suggests that the Sino-US AI race has been advancing since Chinese startup DeepSeek launched early last year. Shook the US technology industry and we put China on the map with an AI model that performs as well as the US but is much cheaper.
Experts say the latest AI models, released by startup Z.ai or Zhipu in June and Moonshot in July, are nearly as smart as leading models from OpenAI and Anthropic. Also in July, China Alibaba’s Previewed the Qwen3.8 Max AI model. In April, DeepSeek released previews of its latest V4 model, challenging OpenAI’s GPT and Google’s Gemini.
These constant advances are paying off. Curt Meinhold, a Greensboro, North Carolina-based technology executive, said he is increasingly turning to DeepSeek for tasks such as finding business leads or ways to generate sales.
“At the end of the day, most of us, the vast majority, more than 90 percent, don’t need (Anthropic’s) Mythos or Fable,” said Meinhold, founder of digital legacy platform LilyList. “Like, we don’t need this, we need something good enough.”
US investment bank Goldman Sachs wrote in a July research report that global AI models in particular have reached a “critical stage” for widespread adoption. “agent” artificial intelligence useAI models that require AI models to execute multi-step, complex tasks autonomously are driving demand for cost-effective AI models.
According to analysts including Alex Colville of the Australian Strategic Policy Institute, AI prices are calculated per million tokens for both input and output, so the use of AI “middlemen” significantly increases cost differences.
“You know the Chinese models, I find them pretty close in terms of code and research,” Meinhold said. “If I can pay a handful of cents per million exit tokens instead of $30, $40, or $50, that’s good enough.”
Moonshot’s Kimi K3 is ahead, but Chinese AI models have limits
According to data last month, the five most popular models on OpenRouter, a platform that tracks data across AI models, belonged to China.
Estimates from Sensor Tower, a market intelligence firm, showed Kimi was downloaded more than 930,000 times during the week after K3 was released in July; This was up 200% from the previous week. It was downloaded nearly 86,000 times in the US, an increase of 387%.
K3 became so popular that Moonshot temporarily suspend new subscriptions Following intense demand, it increased its capacity close to its limits.
Still, while Chinese models are emerging as serious rivals for the U.S., they generally lag behind American AI leaders in terms of their full-range capabilities, said Anastasios Angelopoulos, co-founder and CEO of Arena, a platform for evaluating AI systems.
Yasir Atalan of the Center for Strategic and International Studies added that US AI firms are looking for more options to lower prices, such as cheaper alternatives to competitive models.
US Policy May Support Chinese AI Models
Most Chinese AI models are open source; That is, at a time when US companies had pioneering models, anyone could study and improve them. anthropic And OpenAI It is closed source.
Chinese model vendors are expected to leverage open source software to promote their global use and adoption, said Lian Jye Su of technology research and advisory group Omdia.
Mozilla’s Krikorian said China’s open-source models are increasingly becoming nearly as good as the closed systems produced by leading U.S. companies, adding that “open border is becoming more and more Chinese-made.” As the United States focuses on restricting Chinese artificial intelligence models, a group of American technology companies, including Microsoft, Meta and Nvidia, signed an open letter supporting “open” artificial intelligence models, published on Friday.
Some experts say that other US technology policies also sometimes give China the upper hand. China’s Z.ai launched the GLM-5.2 model in mid-June, shortly after the Trump administration imposed export controls on Anthropic’s Fable and Mythos models, keeping them offline for more than two weeks.
“Restricting an American model could create an immediate opportunity for a Chinese competitor,” Arena’s Angelopoulos said.
Chinese AI models aim to expand globally
People and businesses in China rapidly adopted artificial intelligence technologies They became rich with government support. Chinese technology companies such as Huawei and Tencent are also embedding artificial intelligence in devices such as smartphones, artificial intelligence glasses and humanoid robots.
Now China is increasingly trying to dominate artificial intelligence abroad. At a flagship technology summit in Shanghai, Chinese President Xi Jinping advocated for open-source artificial intelligence models and the importance of promoting greater global equality. I am committed to China’s participation Increasing artificial intelligence capabilities, especially in developing countries.
As is true for many other industries in China, intense competition within the country is pushing companies to expand globally. China’s leading AI startups are now raising more funds, including IPOs, to support the effort.
Chelsey Tam of investment research firm Morningstar said both China and the United States will want to “encourage widespread adoption of AI ecosystems while protecting technologies that could materially strengthen their strategic rivals.”
“The competition is no longer just the United States against China; Chinese laboratories are also putting a lot of pressure on each other,” Angelopoulos said. he said.
However, as in the US, large investments in artificial intelligence have also raised concerns about the sustainability of these Chinese startups. Za.ai, for example, reported that its revenue rose 132% to 724 million yuan ($107 million) last year, but net losses rose 60% to 4.7 billion yuan ($694 million).
Krikorian said that for now, Chinese AI models will continue to advance, including in the United States. “I would strongly recommend that anyone dealing with a serious AI payload at least consider this,” Krikorian said.
___
O’Brien reported from Providence, Rhode Island. AP Business Writer Kelvin Chan He contributed to this report in Toronto.
