Community Analytics
Model popularity data from the HuggingFace Hub, refreshed weekly
Loading popularity data...
// READOUT
This page tracks what the local-AI community actually runs — download and like counts from the HuggingFace Hub, refreshed weekly. Use it to spot momentum, but remember popularity isn't fit: read how to read these numbers before you pick.
01 Model Popularity Dashboard
02 How to Read These Numbers
These charts use download and like counts from the HuggingFace Hub — but raw popularity needs a little interpretation before you act on it.
Last-30-day pulls — a strong usage signal, but inflatable. CI pipelines, mirrors, and automated runs all count, so a spike isn't always human adoption.
Cumulative manual endorsements — a slower, more deliberate signal of approval than downloads. Harder to game, but it lags new releases.
Fresh releases spike then settle; long-stable models show steady numbers. A model at the top of “trending” hasn't necessarily proven itself yet.
Gated, private, and renamed repos undercount. And popularity ≠ quality or fit — the most-downloaded model may not run on your VRAM or suit your task.
Popularity is a starting point, not a verdict. To turn it into a pick that fits your hardware, run the Model Targeting System or the Stack Builder.
03 Data Sources & Methodology
Model Downloads & Likes
Sourced from the HuggingFace Hub and refreshed weekly by our build pipeline. These numbers represent last-30-day downloads and cumulative community likes.
Hardware Stack Builder
Our Hardware Stack Builder lets you select your GPU or Apple Silicon, see which models fit in your VRAM, estimate inference speeds, and save configurations locally in your browser — never sent to any server.