Hugging Face Scraper — Models, Datasets, Spaces & Creator Leads

Export the richest Hugging Face Hub dataset on Apify. Pull models, datasets and Spaces with the metrics that matter — downloads, likes, trending score, license, task, library, parameter count, evaluation results, linked Spaces and file/storage sizes — and resolve every repo's author or organization to a profile with website, socials and a contact lead. Built on the official Hugging Face Hub API: keyless, no browser, no login.

Why this Hugging Face scraper?

Other Hub scrapers do one thing — models or datasets or papers — and stop at the listing fields. This actor unifies the whole Hub, reads each repo's full detail record, and adds the creator-lead layer none of them have.

Data	Typical scrapers	This actor
Models	✅	✅
Datasets & Spaces in the same run	❌	✅
Downloads, likes & trending score	✅	✅
License, task, library, languages, base models	partial	✅
Parameter count, file count & storage size	❌	✅
Evaluation results (model-index)	❌	✅
Linked Spaces / models	❌	✅
Author / org profile (stats, website, socials)	❌	✅ opt-in
Contact email lead (website crawl)	❌	✅ opt-in
Popularity & lead scores (0–100)	❌	✅
Monitor mode — only new releases	❌	✅
Start fee	sometimes	❌ never

Use cases

• AI/ML competitive intelligence — track a competitor's or organization's models, datasets and Spaces: how many downloads and likes, which tasks and licenses, how fast they ship. Schedule recurring runs to watch momentum.
• Model & dataset discovery — find the top models for a task (text-generation, text-to-image, automatic-speech-recognition), filter by library, license, language or minimum downloads, and rank by popularity.
• Lead generation for AI infra & MLOps — turn on author profiles to collect the org/lab behind each model, with website, socials and a contact email — ideal for selling GPUs, inference, eval, data-labeling and tooling into AI builders.
• Recruiting & talent sourcing — surface the authors publishing the most-downloaded models in a domain, with their profile and links.
• Trend & release monitoring — monitor mode emits only newly released models/datasets/Spaces, so you get a clean feed of what just dropped in a field or from an org.

How to use

1. Sign up for Apify — the free plan is enough to try this actor.
2. Open the Hugging Face Scraper, enter search queries, repo URLs or author names (or leave them empty to pull the most-downloaded models), pick your content types, and click Start.
3. Watch results stream into the dataset table as each repo is parsed.
4. Export as JSON, CSV, Excel, XML, or RSS — or pull results programmatically via the Apify API.

Input

{
    "searchQueries": ["llm", "text-to-image"],
    "contentTypes": ["models", "datasets"],
    "sortBy": "downloads",
    "pipelineTag": "text-generation",
    "minDownloads": 1000,
    "includeFullMetadata": true,
    "includeAuthorProfiles": true,
    "enrichContactEmails": true,
    "maxResults": 100,
    "maxResultsPerQuery": 60,
    "monitorMode": false
}

• searchQueries — free-text searches run against each selected content type.
• contentTypes — any of models, datasets, spaces (default models).
• repoUrls — direct model/dataset/Space pages or bare ids (e.g. meta-llama/Llama-3.1-8B-Instruct).
• authors — usernames or organizations to pull a profile + lead (e.g. meta-llama, stabilityai).
• startUrls — optional mixed list of huggingface.co URLs, auto-classified.
• browseHub (default false) — also browse the top of the Hub; auto-enabled when no other input is given.
• sortBy (default downloads) — downloads, likes, trendingScore, lastModified or createdAt.
• pipelineTag / libraryName / language / license / authorFilter / minDownloads / minLikes / excludeGated — filters.
• includeFullMetadata (default true) — fetch each repo's detail (params, files, eval results, linked Spaces). Turn OFF for a fast, cheap listing-only crawl.
• includeAuthorProfiles (default false) — resolve each repo's author/org to a profile and attach a lead.
• enrichContactEmails (default false) — crawl the author's website for a contact email.
• hfToken (optional) — a read-only Hugging Face token for higher rate limits; never required.
• monitorMode (default false) — emit only repos/authors not seen in previous runs (see Monitoring).

Output

One record per repo (type: "model", "dataset" or "space"), plus optional author records:

{
    "type": "model",
    "source": "huggingface",
    "id": "meta-llama/Llama-3.1-8B-Instruct",
    "author": "meta-llama",
    "name": "Llama-3.1-8B-Instruct",
    "url": "https://huggingface.co/meta-llama/Llama-3.1-8B-Instruct",
    "pipelineTag": "text-generation",
    "libraryName": "transformers",
    "tags": ["facebook", "llama", "conversational"],
    "languages": ["en", "de", "fr"],
    "license": "llama3.1",
    "baseModels": ["meta-llama/Llama-3.1-8B"],
    "downloads": 6562137,
    "likes": 6090,
    "trendingScore": 12,
    "parameters": 8030261248,
    "filesCount": 17,
    "storageBytes": 16060549824,
    "linkedSpaces": ["huggingface-projects/llama-3.1-8b-instruct"],
    "evalResults": [{ "task": "text-generation", "dataset": "MMLU", "metric": "acc", "value": 0.73 }],
    "createdAt": "2024-07-18T16:54:00.000Z",
    "lastModified": "2024-09-25T17:00:00.000Z",
    "authorProfile": {
        "name": "meta-llama",
        "url": "https://huggingface.co/meta-llama",
        "website": "https://ai.meta.com",
        "socials": { "github": "https://github.com/meta-llama" },
        "contactEmails": null,
        "leadScore": 47
    },
    "popularityScore": 90,
    "scrapedAt": "2026-06-16T12:00:00.000Z"
}

Fields are null only when the data genuinely doesn't exist. Empty results are reported as a successful run.

Monitoring — get only new releases

Turn on monitorMode and the actor remembers the repo/author ids it has already returned (in a named key-value store) and emits only new ones next time. Point it at a search, an organization, or the whole Hub sorted by createdAt and you get a clean feed of newly released models, datasets and Spaces.

It is orthogonal to Apify Schedules: the schedule decides when a run happens, monitor mode decides what's new, so the two compose cleanly. Use a distinct monitorStoreName per watch to track them independently.

Automate & schedule

Run this actor on autopilot and pull results into your own stack:

• Apify API — start runs, fetch datasets, and manage schedules over REST.
• apify-client for JavaScript and apify-client for Python — official SDKs.
• Schedules — run it hourly/daily/weekly to keep downloads, likes and release data fresh.
• Webhooks — trigger downstream actions the moment a run finishes.

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: 'MY_APIFY_TOKEN' });

const run = await client.actor('scrapesage/hugging-face-scraper').call({
    searchQueries: ['text-to-image'],
    contentTypes: ['models'],
    sortBy: 'trendingScore',
    includeAuthorProfiles: true,
    maxResults: 200,
});

const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(`Got ${items.length} repos`);

Integrate with any app

Connect the dataset to 5,000+ apps — no code required:

• Make — multi-step automation scenarios.
• Zapier — push new models or leads straight into your CRM or sheet.
• Slack — get notified when a monitored org ships a new model.
• Google Drive / Sheets — auto-export every run to a spreadsheet.
• Airbyte — pipe results into your data warehouse.
• GitHub — trigger runs from commits or releases.

Use with AI assistants (MCP)

The output is clean, LLM-ready JSON. Call this actor from Claude, ChatGPT, or any agent framework through the Apify MCP server — ask your assistant to "find the top 20 text-to-image models by downloads, with their license and the org behind each" and let it run this scraper for you.

Agent-ready: autonomous payments (x402 & Skyfire)

This actor is agent-ready — AI agents can discover it, run it, and pay for it autonomously, with no Apify account and no human in the loop. It uses pay-per-event pricing and limited permissions, so it qualifies for Apify's agentic-payment standards:

• x402 — an open, HTTP-native payment protocol. Agents pay per run in USDC on the Base network directly through the Apify MCP server — no account, no API key.
• Skyfire — agent-to-service payments for fully autonomous AI-agent workflows.

Building an AI agent, MCP tool, or autonomous data pipeline? This scraper is ready to plug in and pay as it goes.

More scrapers from scrapesage

Build a complete AI, product & competitive-intelligence stack:

• Product Hunt Scraper — product launches, makers and leads.
• Google Ads Transparency Scraper — who's advertising what on Google.
• Facebook Ad Library Scraper — Meta & Instagram competitor ad intelligence.
• LinkedIn Ad Library Scraper — B2B competitor ads & creatives.
• Multi-ATS Job Scraper — jobs from Greenhouse, Lever, Ashby, Workday & more (source AI/ML talent).
• Eventbrite Scraper — events, ticket prices and organizer leads.
• SAM.gov Scraper — US federal contract opportunities & contacts.

Tips

• Default run: leave all source inputs empty to pull the most-downloaded models — or set sortBy: "createdAt" with monitorMode for a "what just launched" feed.
• Multiple types at once: set contentTypes: ["models", "datasets", "spaces"] and each search/browse runs across all three.
• Cost control: turn includeFullMetadata OFF for a fast, cheap listing-only crawl (still includes downloads, likes, license and tags), then run full metadata only on the repos you keep.
• Leads: includeAuthorProfiles + enrichContactEmails resolves the org/lab behind each model and crawls its site for a contact email.
• Rate limits: the Hub limits anonymous traffic per IP; a fresh proxy session is used per request, and you can add a free hfToken for extra headroom on big runs.

FAQ

Where does the data come from? The official Hugging Face Hub API (huggingface.co/api/*) — no API key required. An optional read-only token just raises rate limits.

Can I scrape models, datasets and Spaces together? Yes — set contentTypes to any combination and each search or browse runs across all of them.

How do I get the organization behind a model? Turn on includeAuthorProfiles to resolve each repo's author/org to a profile (stats, website, socials), and enrichContactEmails to crawl that site for a contact email.

Can I monitor new model releases? Yes — turn on monitorMode (optionally with a Schedule) and sort by createdAt to emit only repos you haven't seen before.

Can I export to Google Sheets, CSV, or Excel? Yes — one click in the dataset view, or automatically on every run via the Google Drive integration.

Is scraping this legal? This actor uses Hugging Face's public API and collects publicly available data only. You are responsible for using the data in compliance with applicable laws (GDPR/CCPA for personal data) and Hugging Face's terms.

Need help?

Open an issue on the actor's Issues tab, or visit the Apify help center. Feature requests are welcome — this actor is actively maintained.