Jin Daily AI Trivia – China’s Open-Source AI Model Download Takeover

Jin Daily AI Trivia – China’s Open-Source AI Model Download Takeover

In the second half of this year, downloads of Chinese open-source large language models (LLMs) surpassed those from the US for the first time.

According to a16z (a major US VC), 80% of Bay Area startups are now using Chinese LLMs for secondary development.

In my 5 sen: Meta—who kicked off the whole open-source trend with Llama—stumbled hard with Llama-4. Meanwhile, the closed-source strategies of Google, OpenAI, Anthropic, and Grok have only accelerated the rise and adoption of Chinese models.

Open Source = Strategic Advantage Open-sourcing isn’t just about code sharing. For China, it’s policy, soft power, and industrial muscle. Faster iteration, bigger local ecosystems, and direct government backing mean Chinese open models are improving quickly—and getting embedded into real products worldwide.

Now the question: Mistral has strong backing from European governments too, but they’ve stopped releasing big open-weight models for a while. So what’s the key difference between the ecosystems in China and Europe?