Hugging Face Scraper — Models, Datasets & Spaces
Pricing
from $0.24 / 1,000 repos
Hugging Face Scraper — Models, Datasets & Spaces
Search the Hugging Face Hub for models, datasets and Spaces: downloads, likes, trending score, task, library, license, base model, languages and arXiv papers. Filter by author, task or library, sort by downloads or trending, and track new repos and download changes between runs.
Pricing
from $0.24 / 1,000 repos
Rating
0.0
(0)
Developer
Northbell
Maintained by CommunityActor stats
0
Bookmarked
2
Total users
1
Monthly active users
2 hours ago
Last modified
Categories
Share
List models, datasets or Spaces from the Hugging Face Hub as clean JSON — by search term, author, task or library, sorted by downloads, likes, trending, newest or recently updated — with downloads, likes, trending score, task, library, license, base model, languages, training datasets and arXiv papers. Schedule a search and each run shows new repos and downloads and likes gained since the last one.
What you get
| Field | What it is |
|---|---|
type, id, owner, name, url | the repo |
downloads, likes, trendingScore | popularity (downloads = last 30 days, as the Hub counts them) |
task, library, sdk | what it does and runs on |
license, languages, datasets, baseModels, arxiv | from the repo's tags |
gated, createdAt, lastModified, tags, description (datasets) | details |
search, rank, isNewSinceLastRun, downloadsChange, likesChange | position in your list, and changes since your previous run |
Fixed 22 Oct 2026: repos below the part of the list your previous run read (it stopped at max results) counted as new, so with Only new a second run could return the next 100 older repos as new and charge for them. New now means it entered the ranks the previous run read, or was created after that run. The ranks counted are the top max results only, so an older repo that drifts up the downloads order from further down is not new either. With Only new and nothing new, the run now stops a little past the ranks it read last time; before, it read every page of the search and could run into Hugging Face's limit of 500 requests per 5 minutes (HTTP 429).
Use it for
- Model landscape tracking — every new fine-tune of a base model, or the top GGUF builds, every week
- AI market research — which organizations and tasks get the downloads
- Dataset discovery — datasets for a topic with license and language
- Trending watch — the trending Spaces or models each morning
Input
Measured on October 7, 2026: 1,200 text-generation models for "qwen" plus one model by ID in 1.3 seconds. Data comes from the public Hugging Face Hub API — no login, no token. The Hub allows 500 requests per 5 minutes; one request returns up to 1,000 repos.
Pricing
$0.30 per 1,000 repos, plus $0.01 per run. Error rows are free. The $0.01 run-start fee is the one charge every run pays, even one that returns nothing.
For AI agents
{"kind": "spaces", "sort": "trending", "maxResults": 20} returns today's 20 trending Spaces. Rows have type: model, dataset, space or error.
