Hugging Face published its biannual analysis of the open-model ecosystem covering January to August 2026. The raw growth is steady: public model repositories went from 2.43 to 2.96 million, datasets from 711,000 to 1 million, Spaces from 1.00 to 1.44 million. The distribution under those totals is what the report insists on reading first — roughly 85.6% of models have fewer than 200 lifetime downloads, and 1.5% of repositories account for 99.2% of all downloads.
The structural finding is about size strategy. 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, with China's monthly ceiling running between 754B and 2.78 trillion parameters while US models stayed under 130B in five of seven months. Several Chinese labs skipped the old progression of starting small and working up, and the report gives two reasons: building large stopped being a differentiator, and a lab no longer needs to ship a small model to be reachable because the community's quantization layer makes a large one runnable within days. That leaves the size profile as a statement of intent rather than capability — a frontier-only portfolio bets on benchmark position and API demand, a full-spectrum one bids to become the family developers standardise on.
Sovereign and open-weight strategy is the frame in Mistral raising €3 billion around a sovereign open-weight stack, and the AI hub tracks the release economics.
The data covers Hugging Face's own hub, so it measures publication and download behaviour there rather than deployment anywhere. The condition to watch is whether the quantization dependency is tracked as infrastructure, because a release strategy that assumes volunteers will make a trillion-parameter model runnable has a single point of failure outside the lab.