What this dashboard shows
Hugging Face is the largest public hub for open AI models, datasets and demo apps, with well over a million models from companies such as Google, Meta, Microsoft, Alibaba (Qwen), Mistral and DeepSeek, as well as universities and individual developers. This dashboard turns the Hub's public data into an at-a-glance view of what is popular and what is new. It loads live data straight from the Hugging Face Hub API in your browser, with no login and no QuickKit server in between, and refreshes every five minutes while the tab is open.
The views
- Trending ranks models by Hugging Face's trending score, which reflects recent likes and activity, the best signal of what people are talking about this week.
- Most downloaded ranks by downloads over the last 30 days, showing the models that actually run in production and in tutorials, often small embedding and speech models rather than the latest headline release.
- Most liked shows all-time favourites by likes.
- New launches lists trending models created in the last 30 days, newest first, a quick way to catch fresh releases with real traction.
- Spaces shows trending demo apps, with their emoji, colours, SDK and likes.
- Organisations combines the top downloaded, trending and liked lists into a leaderboard of the most influential publishers.
- Compare puts up to four models side by side with bars for each metric and the leader highlighted.
- Watchlist keeps the models you star in this browser.
Metrics explained
Downloads count file requests over the last 30 days as reported by the Hub; the per-day figure is that number divided by 30. All-time downloads is the cumulative total. Likes are explicit votes by Hub users. Growth since your last visit is calculated locally: each time you open the dashboard, your browser stores a small snapshot of the numbers it saw, and the next visit compares against it, so the indicator and the small sparkline become more informative the more often you check. Charts show the top models by downloads and the mix of tasks in the current list. Download counts are a rough popularity signal, not a quality measure: automated pipelines can inflate them, and gated or very new models may be under-counted.
Filters, sharing and export
Filter by task on the server, and by library, license and free-text search in the browser. Click an organisation in the leaderboard to see its models. The address bar keeps your view and filters, so Copy link shares exactly what you see, and Export CSV downloads the current table for a spreadsheet.
Privacy and caching
This is the only QuickKit page that contacts another website: your browser requests public data from huggingface.co, which will see your IP address like any website you visit. No cookies or credentials are sent. Responses are cached in local storage for five minutes, so reloading is instant and the last data stays visible if the API is unreachable or rate-limited.