State of Open Models: Summer 2026 Observations

Hugging Face's biannual analysis of the open model ecosystem from January to August 2026 shows explosive growth: public model repositories grew from 2.43 to 2.96 million, datasets from 711,000 to 1 million, and Spaces from 1.00 to 1.44 million. The distribution remains extremely skewed: roughly 85.6% of models have fewer than 200 lifetime downloads, while 1.5% of repositories account for 99.2% of all downloads.

The frontier is moving fast, and the report highlights that several Chinese labs skipped the traditional small-to-large progression. In almost every month of 2026, the largest open model from a Chinese lab was larger than anything an American lab released of its own. China's monthly parameter ceiling ranged from 754B to 2.78 trillion, while America's own ceiling stayed under 130B in five of seven months, with exceptions being NVIDIA's Nemotron 3 Ultra (561B) and Thinking Machines Lab's Inkling. The report divides labs into two camps: Moonshot, MiniMax, Xiaomi, and Z.ai publish almost nothing below 70B, whereas Tencent and Alibaba Qwen cover the full range from under 1B upward.

Building large is no longer a differentiator—Xiaomi and Meituan both cleared a trillion parameters this year despite being relatively unknown in open weights a year ago. The community's quantization layer makes large models runnable within days, so a frontier-only portfolio stakes on benchmark position and API demand, while a full-spectrum portfolio aims to become the standard family. The US is not absent: AMD and NVIDIA each released over 200 new model repositories, far ahead of others, with LiquidAI third at around 100, showing that hardware vendors use open models to sell chips.