HuggingFace Models Scraper

Scrape AI/ML models from HuggingFace — the world's largest repository of open-source machine learning models.

Extracts structured data including model ID, download counts, likes, task type (pipeline tag), ML library, tags, gated status, and timestamps. Powered by the official HuggingFace API — no web scraping, no rate-limit surprises.

What it does

• Fetches models from the HuggingFace public API with full metadata
• Supports filtering by keyword search, author/organization, task type, and library
• Supports sorting by downloads, likes, or date
• Paginates automatically up to your specified limit (up to 10,000 models)

Use cases

• AI research: Track which models are trending by downloads or likes
• Competitive intelligence: Monitor what models a specific organization has published
• Dataset building: Collect model metadata for ML benchmarks or surveys
• Lead generation: Find organizations actively publishing models in your domain
• Content & newsletters: Curate the most popular or newest models by task type

Input

Field	Type	Default	Description
search	string	—	Keyword search to filter models
author	string	—	Filter by author or organization (e.g. meta-llama)
pipelineTag	string	—	Filter by task type (e.g. text-generation, image-classification)
libraryName	string	—	Filter by ML library (e.g. transformers, diffusers)
sort	select	downloads	Sort by: downloads, likes, createdAt, lastModified
maxModels	integer	100	Max models to return (1–10,000)

Example input

{
  "search": "llama",
  "pipelineTag": "text-generation",
  "sort": "downloads",
  "maxModels": 50
}

Output

One record per model saved to the default dataset.

Field	Type	Description
modelId	string	Full model ID (e.g. meta-llama/Llama-3.1-8B-Instruct)
author	string|null	Author or organization name
downloads	number|null	Total download count
likes	number|null	Like count
pipelineTag	string|null	Task type (e.g. text-generation)
libraryName	string|null	ML library (e.g. transformers)
tags	string[]	All tags including datasets, licenses, frameworks
gated	boolean|null	Whether model access requires approval
createdAt	string|null	Creation date (ISO 8601)
lastModified	string|null	Last modified date (ISO 8601)
url	string	Direct URL to the model page
scrapedAt	string	Timestamp when this record was scraped

Example output

{
  "modelId": "sentence-transformers/all-MiniLM-L6-v2",
  "author": "sentence-transformers",
  "downloads": 208493944,
  "likes": 4598,
  "pipelineTag": "sentence-similarity",
  "libraryName": "sentence-transformers",
  "tags": ["sentence-transformers", "pytorch", "onnx", "license:apache-2.0"],
  "gated": false,
  "createdAt": "2022-03-02T23:29:05.000Z",
  "lastModified": "2025-03-06T13:37:44.000Z",
  "url": "https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2",
  "scrapedAt": "2026-03-22T03:46:43.767Z"
}

Features

• Official API — Uses the HuggingFace REST API directly; no fragile HTML parsing
• Automatic pagination — Fetches all pages until your limit is reached
• Polite rate limiting — 500ms delay between API calls
• Robust input validation — Clear error messages for invalid inputs

Notes

• Results are limited to public models only; private models are not accessible
• The gated field indicates whether a model requires access approval from the author
• HuggingFace API does not support combining search with all sort orders equally; downloads sort works best for broad searches
• Download counts are 30-day rolling totals as reported by HuggingFace

Support

Found a bug or have a feature request? Please open an issue or contact the author through the Apify platform.