Open-weight AI, measured
AI Observatory
Where open-weight AI is built, adopted, and forked — measured continuously from Hugging Face and GitHub by the Observatory's own pipeline.
Map
Where tracked organizations sit on the globe.
Graph
Organizations, models, and repositories as one network.
Models
Every tracked model, searchable, with cards and lineage.
Benchmarks
External benchmark leaderboards across modalities.
Evaluations
Self-reported evaluation results from model cards.
Analytics
Downloads, derivatives, and quality by country over time.
Ecosystem
Active forks and how code traction bridges into adoption.
Methodology
expandcollapse
Every number on this site comes from the Observatory's own pipeline against primary sources: the Hugging Face and GitHub public APIs. Model adoption uses Hugging Face downloads; code activity uses repository stars, forks, and recent fork pushes; geography is resolved per organization and normalized to countries.
No third-party datasets are ingested. Counts reflect the tracked corpus, not all of Hugging Face; collection timestamps are shown wherever the data is time-sensitive.