A landmark shift just occurred in global AI. According to a joint MIT–Hugging Face study, Chinese open-source models now account for 17.0 % of all downloads on Hugging Face, surpassing the United States’ 15.8 % for the first time. Europe sits at 12.4 %, with the rest of the world sharing the remainder.
The leaders are unmistakable: DeepSeek’s V3 series and Alibaba’s Qwen family dominate the most-downloaded open-weight rankings. Together, these two organizations alone represent nearly 10 % of global downloads. Add ByteDance, 01.AI, Zhipu, Moonshot, and dozens of smaller labs, and China’s presence becomes overwhelming.
The strategy is deliberate. Beijing actively encourages open-weight releases through grants, tax incentives, and regulatory carve-outs that allow Chinese labs to publish full model weights while many Western counterparts are forced to keep frontier systems closed.
The result: China now fields more top-tier open models in a single year than the rest of the world combined. Chinese models particularly dominate multimodal tasks, video generation, image understanding, and non-English performance, with Qwen-VL and DeepSeek derivatives regularly topping independent leaderboards.
The pattern mirrors other industries. While American labs concentrate resources on closed, frontier-scale systems, Chinese organizations flood the market with high-quality, fully open models that run efficiently on modest hardware. DeepSeek-V3-236B, for example, outperforms every other open model on coding and math while fitting on a single 8×H100 node with quantization.
Qwen 2.5-72B has become the default choice for enterprises and governments requiring on-premise deployment for data-sovereignty reasons. These models are not just competitive; they are cheaper to host, easier to customize, and licensed under permissive terms.
Real-world adoption reflects the shift:
- A large share of new enterprise deployments in Southeast Asia, the Middle East, and Latin America now choose Chinese open-weight bases over Llama or Mistral derivatives.
- Independent developers in India, Brazil, Indonesia, and Eastern Europe increasingly fork Chinese models because they ship with superior multilingual alignment out of the box.
- In open video-generation leaderboards, seven of the top ten models are Chinese.
This is classic volume strategy: sacrifice short-term licensing revenue for ecosystem dominance. Every startup, university, factory, or government that builds on Qwen or DeepSeek becomes part of China’s extended distribution network, feeding usage data back and creating a moat that closed models cannot replicate.
It’s the same playbook that turned Huawei into a telecom giant, Xiaomi into a smartphone powerhouse, and CATL into the world’s largest battery maker.
The West still owns the absolute frontier; closed models from OpenAI, Anthropic, and Google remain unmatched in raw capability. But the frontier represents only a tiny fraction of actual economic value. The vast majority of real-world use cases (regional chatbots, factory automation, legal processing, education tools) don’t need superintelligence; they need reliable, affordable, private, and customizable systems.
And that segment just quietly changed hands.
For the first time in the AI era, the United States no longer leads open-source mindshare. The long-term consequences for developer ecosystems, technological sovereignty, and the global geography of innovation are only beginning to unfold.
Also read:
- OpenAI's High-Wire Act: Chasing $200 Billion in Revenue While Staring Down a $207 Billion Funding Abyss
- We Need More Weird Streaming Services, and Letterboxd Just Delivered the Strangest One Yet
- TikTok Will Start Handing Out Digital Medals If You Stop Dooming at 3 A.M.
Thank you!
Author: Slava Vasipenok
Founder and CEO of QUASA (quasa.io) — the world's first remote work platform with payments in cryptocurrency.
Innovative entrepreneur with over 20 years of experience in IT, fintech, and blockchain. Specializes in decentralized solutions for freelancing, helping to overcome the barriers of traditional finance, especially in developing regions.
