Chinese AI labs bypass small models, ship trillion-parameter open weights in 2026
GoKawiil's summer analysis of Hugging Face data shows public model repositories rose from 2.43M to 2.96M between January and August 2026, with datasets and Spaces also expanding sharply. Despite that growth, download activity remains concentrated in a tiny fraction of repositories, and Chinese labs including Moonshot, MiniMax, Xiaomi and Z.ai are releasing massive models, some exceeding a trillion parameters, without first publishing smaller versions.
GoKawiil's interpretation of the reporting above, not reported fact.
This marks a shift from the traditional pattern where labs built credibility with small, widely usable models before scaling up. Now, community-driven quantization tools let developers shrink huge open models for local use, so labs can skip the small-model stage entirely and still reach a broad audience, changing how open-weight ecosystems form and who dominates them.
- HF repositories grew nearly 22% (2.43M to 2.96M models) from January to August 2026, with 85.6% of models getting fewer than 200 downloads.
- Chinese labs' top monthly model sizes (754B–2.78T parameters) consistently outpaced U.S. releases, which mostly stayed under 130B.
- Xiaomi, Ant Group and Meituan surpassed one trillion parameters in 2026 despite limited prior visibility in open-weight releases.
Apple Mac mini (M4, 24GB unified memory) — With open-weight models pouring onto Hugging Face by the millions, the fun part is actually running them yourself. The M4 Mac mini's unified memory makes it a compact, quiet local-inference box for tools like Ollama or LM Studio, so you can pull down the latest open release and test it the day it drops.
See Apple Mac mini (M4, 24GB unified memory) on Amazon → Affiliate link — we may earn a commission on purchases, at no extra cost to you. Product picked by AI based on this article; it is not a tested recommendation.Source: huggingface.co, 2026-09-01
Published there as: “State of Open Models: Summer 2026 Observations”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.