Domestic open source model has been downloaded more than 10 billion times globally: from "being noticed" to "being used on a large scale"

The global cumulative downloads of domestic open source models have officially exceeded 10 billion times. What is more noteworthy than this number is that the share of Chinese models in the global consumption of large model Tokens continues to be highlighted - they are not just "downloaded", but are actually deployed and called, and are deeply embedded in the application stacks of global developers.

10 billion downloads is proof of "scale"

Downloads are the first-level indicator of open source influence. Breaking through 10 billion means that domestic open source models represented by DeepSeek, Qwen, GLM, etc. have become one of the default technology options for developers around the world. "Being noticed" of this magnitude is influence in itself.

Tokens consumption is the "real influence"

Compared with the download volume, Tokens consumption reflects the actual scale of the model being called. The prominent influence of domestic models in Tokens consumption shows that they have moved from "collection and early adopters" to "production-level applications" - and high usage will in turn drive the growth of inference computing power and commercial API demand, forming a positive cycle.

Open source as an amplifier of globalization

This milestone further verifies the effectiveness of "open source" as China's AI globalization strategy: opening up weights in exchange for a global developer ecosystem, thereby affecting model usage habits and standard discourse rights. Of course, there are always geographical and compliance variables when going overseas. The larger the scale, the more variables need to be carefully managed, which also determines whether "downloads" can be transformed into sustainable ecological and commercial value.

Several directions worth tracking in the future:

1. Composition Breakdown: The composition of 10 billion downloads (which models contribute the most) and the growth trend.
2. Share source: The specific share and growth source of domestic models in Tokens consumption.
3. Business Feedback: Feedback from high usage on computing power costs and business models.
4. Geographical Variables: Geographical and compliance variables faced by open source going global.