A set of download data for open-source models reveals the most authentic choices of developers worldwide.

Since July of this year, an interesting change has emerged—the cumulative downloads of Chinese open-source models have surpassed those of the United States for the first time, while European models continue to grow steadily. This reflects not only technological competition but also the actual voting of the global developer community.

The data is very straightforward: Qwen's downloads have already exceeded the total of all other models, and this overwhelming advantage indicates something. Llama remains in the second tier, with Mistral and DeepSeek standing shoulder to shoulder in third place, while OpenAI's open-source models have almost no presence in this track.

This shift in the landscape is the most genuine reflection of "voting with feet" in the open-source ecosystem. Developers' choices are pointing the market in a certain direction.