Kimi K3: The Chinese AI model that challenges OpenAI and Anthropic

  • Moonshot AI launches Kimi K3, a 2,8 trillion parameter model that matches US leaders.
  • The model is open source and can be downloaded for free, which threatens the business model of OpenAI and Anthropic.
  • Accusations of industrial espionage and chip restrictions are not slowing China's advance.
  • The impact in Europe: Spanish companies could adopt cheaper Chinese models, defying the AI ​​Act regulation.

Kimi K3 AI model

Chinese artificial intelligence is making waves once again. Moonshot AI has unveiled Kimi K3 , a massive language model with 2,8 trillion parameters that, according to initial assessments, matches and even surpasses the most advanced systems from OpenAI and Anthropic in some tasks. Most striking is that it's an open-source model: anyone can download its weights and run it without paying per query, something its American rivals don't offer.

The launch has caused a seismic shift in Silicon Valley and reignited the debate about the effectiveness of Washington's export restrictions on chips. Kimi K3 demonstrates that China can innovate even with limited hardware , and its impact is already being felt in the stock markets, where Nvidia's shares fell after the announcement. For Europe, the arrival of such a powerful and free model poses a dilemma: Spanish companies could adopt it to reduce costs, but that clashes with the EU's efforts to regulate AI with the AI ​​Act.

Kimi K3
Related article:
Kimi K3: the new Chinese AI model that challenges OpenAI and Anthropic

An open-source model that challenges US dominance

Kimi K3 AI model

Kimi K3 is not just another Chinese model. With 2,8 trillion parameters and an expert-mix architecture that activates only 16 of its 896 internal specialists per task, it achieves elite performance at a cost up to ten times lower than its competitors. Furthermore, it supports one million context tokens, allowing it to process entire books in a single run. Moonshot has published the weights on Hugging Face, enabling developers worldwide to download, modify, and host the model without relying on external servers.

This openness contrasts sharply with the strategy of OpenAI and Anthropic, which keep their most powerful models under lock and key and charge for access. The threat to their valuations is real : both companies are preparing IPOs with valuations approaching one trillion dollars, but the existence of a free and comparable alternative could cut those figures. Analysts like Salim Ismail are already talking about a "Sputnik moment" for AI.

Espionage accusations and Washington's response

The success of the Kimi K3 has not gone unnoticed at the White House. US officials have accused Moonshot of using distillation techniques to extract capabilities from models like Anthropic's Claude, a practice they consider intellectual property theft. Treasury Secretary Scott Bessent has threatened sanctions if the industrial espionage is confirmed. However, analysts like Wei Sun of Counterpoint Research point out that the timeline doesn't add up: Moonshot wouldn't have had enough time to distill the newer models before the K3's release.

For his part, Nvidia CEO Jensen Huang has defended Chinese AI companies. Huang believes that open-source models like Kimi K3 are positive for hardware and will not displace US companies. "Free AI should be good for chips," he stated in an interview with Axios. His comments contrast with the stance of Anthropic, which has called the situation a "national challenge."

What does it mean for Spain and Europe?

The emergence of Kimi K3 has direct implications for the European technology ecosystem. Spanish companies developing AI-based services might be tempted to adopt Chinese models due to their lower cost and open nature, reducing their dependence on US providers. This, however, clashes with the European Union's AI Act, which seeks to impose ethical and transparency criteria on artificial intelligence systems. The difficulty of regulating models that are freely distributed from outside the EU poses a significant regulatory challenge.

Furthermore, the rise of Chinese models could influence the biases and values ​​they incorporate, affecting the information consumed by citizens and businesses. Brussels will have to decide whether to tighten controls or accept a greater diversity of sources in the AI ​​market. Meanwhile, the IPOs of OpenAI and Anthropic will serve as a barometer of investor confidence in US technological leadership.

Kimi K3 is not just another model; it's proof that the gap between Chinese and American AI has shrunk to a matter of months. With its efficient architecture, low cost, and open-source license , Moonshot AI has achieved what seemed impossible: making a cutting-edge system accessible to everyone. Washington's reaction, investor concerns, and the interest from European companies confirm that the artificial intelligence landscape has changed forever. It remains to be seen whether restrictions on chips and accusations of espionage will manage to halt Chinese progress, or whether, as Jensen Huang suggests, open competition will benefit the entire sector.


Add as preferred source in Google