source: Hugging Face Blog: State of Open Models: Summer 2026 Observations
level: business
Hugging Face released its summer 2026 open model observations, covering January through August. Public model repositories grew from 2.43 million to 2.96 million, datasets from 711,000 to 1 million, and Spaces from 1.00 million to 1.44 million. The distribution remains extreme: 85.6% of models have fewer than 200 lifetime downloads, while 1.5% of repositories account for 99.2% of all downloads.
Chinese labs now release the largest open models, with monthly parameter ceilings between 754 billion and 2.78 trillion, while American labs stayed under 130 billion in five of seven months. Attention and adoption diverge sharply: only one repository appears in both the top 25 by downloads and top 25 by likes. Small models under 1 billion parameters take 83% of all-time downloads, while models above 100 billion take just 1%.
Qwen has become the community's base model, with 151,448 derivatives on the Hub, 2.6 times Meta's footprint. Licensing is permissive: 59% of Chinese releases above 20 billion parameters use Apache 2.0, and none use non-commercial restrictions. Coding agents are a growing user base, with Claude Code leading at 44.4% in July, but shares shift rapidly month to month.
why it matters: The data shows where open model value is accumulating, helping developers choose which ecosystems to build on and which models to deploy.
source: Hugging Face Blog: State of Open Models: Summer 2026 Observations