Chinese AI model takes US tech industry by surprise with abilities rivaling Claude and ChatGPT
Another powerful new AI model from China surprised the US tech industry on Friday; the last sign of this Chinese startups Those who publicly release their “open source” AI technologies are making the California giants of AI sweat.
The latest Kimi K3 model from Moonshot, a Beijing-based startup run by a Pink Floyd-loving entrepreneur who earned his PhD in Pittsburgh, looks set to capture the best versions of Anthropic’s Claude and OpenAI’s ChatGPT.
“This could be the biggest release of the year,” said Anastasios Angelopoulos, co-founder and CEO of Arena, a platform for evaluating artificial intelligence systems, and marks a moment when open-source Chinese models surpass closed U.S. models.
Advertising
Advertising
Kimi K3 topped the charts in Arena’s rankings for what it calls “front-end coding ability,” a measure of the AI big language model’s performance. “More results are coming that will continue to show he is at the top of the pack,” Angelopoulos said on social media.
It was probably no coincidence that the K3 was introduced shortly before that. Chinese President Xi Jinping On Friday, he keynotes the country’s annual World Artificial Intelligence Conference in Shanghai.
US-led restrictions blocked china Access to some of the world’s most advanced technologies is being blocked, spurring China’s efforts to build its own know-how and intensifying competition between the world’s two largest economies.
“The development of artificial intelligence should be a symphony of global cooperation, not the solitary performance of a single country,” Xi said at the event. he said.
Chinese artificial intelligence models have made great strides
K3 is the sequel to another major AI model that Chinese startup Zhipu, or Z.ai, released last month. The new flagship GLM-5.2 model is already widely used by software developers around the world, who say it can perform almost as well as the top US models at a lower price.
Advertising
Advertising
Excitement about new Chinese model resembles market-shattering panic that followed Chinese startup DeepSeek Although not everyone finds this justified, the new model is expected to be released in early 2025. Technology analyst Patrick Moorhead said on social media that the reaction to K3 was a “strikingly similar” overreaction to DeepSeek’s announcement last year. This may be good for some parts of the broader AI industry, he said, but poses a revenue problem for Anthropic and OpenAI.
During the conference, which will last until Monday, technology giant Huawei also showcased a new artificial intelligence computing system called Atlas 950 SuperPoD; It’s a sign that China is increasingly amassing the domestic hardware it needs despite U.S. restrictions on imports from chipmakers like Nvidia.
Moonshot didn’t say what hardware it used to build the K3, but the startup is in partnership with Huawei.
The price of using K3 is the highest ever for a Chinese AI model, but it’s still half as expensive as OpenAI’s high-performance GPT-5.6 Sol model, Bank of America research analysts reported Friday.
Advertising
Advertising
US politicians and several major US AI companies, including Anthropic and OpenAI, have accused Chinese AI models of being illegal “distillation” Beijing is using models to extract its own technologies, which it says are “unfounded”.
anthropic In February, they accused DeepSeek, Moonshot and a third China-based AI lab, MiniMax, of participating in campaigns to “illegally acquire Claude’s abilities to improve their own models” using the distillation technique, which “involves training a less capable model on the outputs of a more powerful model.”
Anthropic said distillation could be a legitimate way to train AI systems, but it’s a problem if competitors “use it to source powerful capabilities from other labs at a fraction of the cost and in a fraction of the time it would take to develop independently.”
But it could go either way. San Francisco-based startup Anysphere, maker of the popular coding tool Cursor, has admitted that one of its best products is based on Moonshot’s K2.5 model. Elon Musk’s SpaceX plans to make a deal to buy Cursor for $60 billion later this year.
K3 marks a leap forward for ‘open source’ AI models
Moonshot co-founder and CEO Yang Zhilin received his Ph.D. He is said to have made significant contributions to the field of machine learning at Carnegie Mellon University in 2019 and is known for his love of rock bands such as Pink Floyd.
Advertising
Advertising
Pride among former colleagues at the Pennsylvania school transcends the U.S.-China rivalry.
“What a win for the open source community! It feels like just yesterday Zhilin was graduating from my lab at CMU,” wrote his former mentor Russ Salakhutdinov, who is also the former director of AI research at Apple.
Developers who develop “open source” AI are making key components of the technology accessible for anyone to study, modify, and improve. While supporters say open source practices encourage innovation, critics warn that making powerful AI models publicly available poses safety and security dangers. ___
Associated Press writer Chan Ho-him contributed to this report.
