The Chinese company that claims: getting closer to OpenAI and Anthropic

According to the company, although the model still does not overtake OpenAI’s GPT 5.6 Sol and Anthropic’s Claude Fable 5 in overall performance, it manages to present better results than other advanced models, including GPT 5.5 and Claude Opus 4.8, mainly in programming tasks and running AI agents, according to details published by the company on its website.

According to Moon-Shot, this is the largest model developed so far in China, with 2.8 trillion parameters, a figure that reflects the scope of the neural network on which the model is based. Bank of America’s research firm noted that despite the limitations of China’s access to advanced AI chips and computing power, Kimi K3 illustrates that improvements in the model architecture and the training process still make it possible to achieve a leap in performance.

Beyond the technological announcement, the launch also affects the financial markets. Investors see it as a kind of reconstruction of another “Deepseek moment”, referring to the upheaval caused at the beginning of the year by the company DeepSeek, which presented an advanced model at low costs and raised questions about the huge investments of the American technology giants in artificial intelligence infrastructures. Once again, the fear that Chinese companies are managing to quickly narrow the technological gap led to declines in chip stocks and declines in futures contracts on the Nasdaq.

Gaining momentum outside of China as well

The launch comes at a time when Chinese artificial intelligence models are gaining momentum outside of China as well. Western companies are increasingly examining the use of Chinese models, among other things due to relatively low costs, while in the United States there is increasing discussion surrounding the possibility of limiting the use of Chinese-sourced AI technologies by American companies.

At the same time, Chinese President Xi Jinping participated this week in the country’s central artificial intelligence summit and emphasized Beijing’s ambition to influence the design of the global rules of the game in the field, a step that illustrates the strategic importance China attaches to the AI ​​race.

Moonshot’s announcement was also felt among the local competitors. The stock of Z.ai, one of the leading artificial intelligence companies in China, which only launched a new model last month and received a lot of attention, fell by about 28%. Shares of MiniMax, another Chinese startup developing artificial intelligence models, lost about 16%. The stock of Alibaba, which develops the Qwen series of AI models and is also one of the investors in Mon-Shot, also fell by about 4%, after rising earlier this week following the announcement of its collaboration with Apple in China.

Moonshot AI was founded in 2023 and is located in Beijing. According to publications, the company raised 2 billion dollars last May at a value of more than 20 billion dollars. Among its investors are, as mentioned, the Chinese technology giants Alibaba and Tencent.

Analysts at Bank of America estimated that Kimi K3 raises the bar for the capabilities of artificial intelligence models in China and shifts the pressure to other AI labs in the country to show similar progress. If DeepSeek marked the first turning point in China’s artificial intelligence race, the markets are already treating the launch of Kimi K3 as a signal that this is not a one-time event, but rather a trend whereby Chinese companies continue to narrow the gap with their American competitors.