Chinese labs shipped the largest open model in almost every month of 2026
Hugging Face's summer review puts China's monthly open-weight ceiling at 754B to 2.78 trillion parameters against under 130B for US labs in most months of 2026. Of 178 Chinese releases above 20B, 59% ship under Apache 2.0 and 22% under MIT, with none carrying a non-commercial restriction — these weights are not a licensing business. That permissiveness has made Qwen the community's base model at 151,448 derivatives on the Hub, 2.6x Meta's footprint, so the default open stack developers build on now originates in China.
Source: huggingface.co ↗
In almost every month of 2026, the largest and most performant open model from a Chinese lab was larger than anything an American lab released of its own.
Hugging Face State of Open Models
Why this matters
- → China now defines the open-model baseline developers build on — supply-chain shift away from US labs
- → Frontier models shipped under MIT/Apache 2.0 globally reshape licensing economics toward permissive tiers
- → Downloads ≠ excitement; adoption measured by derivatives and long-tail infrastructure, not launch-week hype
China's open-model dominance