Hugging Face ranks repositories by a trending score that rises with recent likes and downloads. The list on this page is that ranking, read from the public API and frozen at the collection time shown above. It answers one question: which open models are getting attention right now.
What we add
For each model the page shows what you would otherwise open several tabs to find: the task tag, the parameter count from the safetensors metadata, the weight formats in the repository, the licence tag with a plain reading of what it permits, and the two numbers that matter for judging attention, downloads in 30 days and likes.
What we take out
A model with 100 or more likes whose 30-day downloads are below one per 10 likes, 3 or more days after the repository was created, is held back. Likes can be gathered without anyone running the model, and the trending score rewards them. Held-back models are listed in a separate box with the numbers that triggered the rule; they are never deleted.
What the list cannot tell you
Trending is not quality. A conversion of a popular model can outrank the original, and a model distributed mostly through an API will look quieter than it is. Use the list to learn what was released; use the model card and independent benchmarks to decide what to run.