Chinese startup releases AI model that surpassed American counterparts

Chinese AI laboratories are capable of competing with American ones in one of the key technological spheres.

The Beijing startup announced on Friday that it plans to fully open-source the model by the end of the month — meaning anyone will be able to download it for free and customize it for their needs. Kimi K3 has 2.8 trillion parameters, which, the company claims, makes it the largest open AI model in the world, writes The Wall Street Journal.

What are AI system parameters and why are they important

Parameters in an AI system work roughly like neurons in the brain: the more there are, the more knowledge the model can retain, so their number is often used as a conditional indicator of power. Some companies disclose these numbers publicly, others do not. Anthropic, for example, has not publicized the number of parameters of its models, although, according to estimates by third-party researchers, Claude Opus 4.8, released in May, has more than 1.5 trillion.

Moonshot's release came as a surprise to many in Silicon Valley and confirms the view that Chinese laboratories are hot on the heels of American competitors. Shortly before this, the Beijing company Z.AI presented its GLM-5.2 model, which in a number of tests, especially in programming, outperformed several leading American systems.

According to data provided by Moonshot itself, Kimi K3 outperformed Anthropic's Opus 4.8 and OpenAI's GPT 5.5 in most programming tests and autonomous AI agent actions. However, the Chinese development is not yet on par with the best closed American models — Fable 5 and GPT-5.6 Sol.

Jefferies analysts called this release an "unexpected breakthrough" and believe it will drive further innovation across the industry.

In AI, the US outpaces China by "six to nine months"

China's rapid progress in artificial intelligence is explained by Beijing's efforts to achieve technological self-sufficiency in competition with the US for leadership in this field.

At a conference on Wednesday, Tarun Chhabra, head of national security at Anthropic, stated that the US lead over Chinese laboratories in AI models is currently "approximately six to nine months." At the same time, Anthropic previously accused a number of Chinese laboratories, including Moonshot and Z.AI, of using American models to train their own systems.

Chinese AI models are also gaining popularity worldwide because many of them are open source and generally cheaper to operate than their Western counterparts.

Who invested in Moonshot

Among Moonshot's investors are the venture fund HSG and the technology giant Alibaba. The company has already begun preparations for an initial public offering (IPO) in Hong Kong.

Moonshot, according to sources familiar with the situation, is valued at $31.5 billion in its current funding round.

• 0
• 0
• 0
• 0
• 0
• 0
• 0
• 0