🇨🇳HF Hub Grows, Chinese Models Dominate
Chinese labs are setting the pace in model size and downloads
TL;DR
The Hugging Face Model Hub has seen massive growth, with Chinese labs leading in model size and downloads. Chinese models now outsize American counterparts, with Qwen reaching 2,045M downloads compared to Moonshot's 37M.
The Hugging Face Model Hub has seen a surge in models and datasets, with public repositories growing from 2.43 to 2.96 million and datasets from 711,000 to 1 million. Chinese labs are dominating, with Qwen from Alibaba and Tencent leading in both size and downloads. The largest and most performant open model from a Chinese lab is now larger than any American lab's model. This shift is significant for developers relying on large models for complex tasks, as Chinese models are becoming the go-to choice for performance and scale. Key numbers: Chinese models range from under 1B to over 2.78 trillion parameters, while U.S. models top out at 130B.
Key Points
Public model repositories grew from 2.43 to 2.96 million over the period, datasets from 711,000 to 1 million.
The largest Chinese model exceeds any American model, with Qwen reaching 2,045M downloads compared to Moonshot's 37M.
Chinese models range from under 1B to over 2.78 trillion parameters, while U.S. models top out at 130B.
Hardware vendors like AMD and NVIDIA have released over 200 new model repositories, pushing the open-source AI ecosystem.
Chinese labs license their largest models about as permissively as their smallest, with 59% carrying Apache 2.0 and 22% MIT.
Why It Matters
If you're working with large-scale AI models, the shift towards Chinese models like Qwen from Alibaba and Tencent is crucial. These models not only lead in size but also in downloads, indicating a significant shift in the AI model landscape. For instance, Qwen's broader release strategy across model sizes has resulted in 2,045M downloads, compared to Moonshot's 37M. This dominance is reshaping the AI model ecosystem, impacting everything from model selection to performance expectations.
Frequently Asked Questions
Why does this matter?
If you're working with large-scale AI models, the shift towards Chinese models like Qwen from Alibaba and Tencent is crucial. These models not only lead in size but also in downloads, indicating a significant shift in the AI model landscape. For instance, Qwen's broader release strategy across model sizes has resulted in 2,045M downloads, compared to Moonshot's 37M. This dominance is reshaping the AI model ecosystem, impacting everything from model selection to performance expectations.
What happened?
The Hugging Face Model Hub has seen massive growth, with Chinese labs leading in model size and downloads. Chinese models now outsize American counterparts, with Qwen reaching 2,045M downloads compared to Moonshot's 37M.
Comments
Be the first to comment
Enjoyed this article?
Get it daily. 7am. Free. Reads in 5 minutes.
Join 3,436 builders reading daily.