China drove this momentum hard. A year after the release of DeepSec, there are now a group of Chinese open source giants following the same blueprint, including Z.ai (formerly Zhipu), Moonshot, Alibaba’s Qwen, and MiniMax. They’re all racing to release more capable models, and they’re closing in on American rivals at a faster pace than expected.
This matters as the AI ​​hype dies down, and companies shift their focus from busy pilots to deployment and integration, where cheaper and more customizable tools tend to win out. Chinese pricing means developers with limited budgets can experiment more, and open weights mean they can adapt models without asking permission.
A study Researchers from MIT and Hugging Face found that Chinese open-weight models accounted for 17.1% of global AI model downloads during the year ending in August 2025. That narrowly surpassed the U.S. share of 15.86%—the first time China had led in this metric. And Embracing facial data Last month showed that Alibaba’s models, including its Kevin family, are now the most popular consumer-made variants—more than Google and Meta combined.
The open source ideal, though, runs headlong into some harsh realities. Chinese models bear the imprint of China’s content moderation system and are trained to avoid content that conflicts with government policy. And in February, Anthropic Accused Several Chinese laboratories illegally extract capabilities from cloud distillation, a process where you use the output of one model to train another. This is standard industry practice, but top US firms such as OpenAI and Anthropic claim that Chinese companies have used fraudulent methods to do this.