Hugging Face Models Scraper

Track public Hugging Face model metadata, popularity, tags, licenses, and update signals from one clean Apify Actor.

What does Hugging Face Models Scraper do?

Hugging Face Models Scraper collects public model catalog records from Hugging Face and saves normalized rows to an Apify dataset. It is built for users who need repeatable model monitoring without manually browsing model pages. You can search by keyword, filter by author, filter by task pipeline, filter by library, and sort by popularity or recency.

Who is it for?

• 🤖 AI teams tracking model ecosystems and popular checkpoints.
• 📈 Market researchers measuring adoption of open-source models.
• 🧪 MLOps teams monitoring competitor model updates.
• 🧠 OSS intelligence teams watching licenses, tags, and download movement.
• 📰 Analysts building reports about AI model trends.

Why use this actor?

The Hugging Face catalog changes constantly. A scheduled scraper gives you repeatable snapshots, structured exports, and automation-ready data. Instead of copying model names, likes, downloads, tags, and licenses by hand, run this actor and export JSON, CSV, Excel, or connect it to your workflow.

Key features

• 🔎 Search public models by keyword.
• 👤 Limit results to an author or organization.
• 🧩 Filter by pipeline tag such as text-generation.
• 🛠️ Filter by library such as transformers or diffusers.
• 📊 Sort by downloads, likes, trending score, or last modified date.
• 🧾 Optional model-card metadata enrichment.
• 🔁 Pagination support for larger model lists.
• 📦 Clean one-row-per-model dataset output.

What data can you extract?

Field	Description
modelId	Full Hugging Face model ID.
author	Owner namespace or organization.
name	Model name without namespace.
url	Public model URL.
likes	Like count.
downloads	Download count reported by Hugging Face.
trendingScore	Trending score when available.
tags	Public tags for the model.
pipelineTag	Primary task / pipeline.
libraryName	Main library or framework.
license	License parsed from tags or card metadata.
createdAt	Creation timestamp.
lastModified	Last modification timestamp.
private	Private flag when reported.
gated	Gated-access flag when reported.
sha	Latest repository commit SHA.
siblingsCount	Number of repository files when available.
cardData	Optional model-card metadata.

Pricing

This Actor uses Apify pay-per-event pricing. The prices below come from the current Actor pricing configuration. Apify public plans map to Store discount tiers, so the table shows both the user-facing plan context and the pricing tier name. The final price shown in Apify depends on the user account plan and any custom agreement.

Event	What is charged	Price
apify-actor-start	Charged when the Actor starts running. Number of events charged depends on Actor memory (one event per GB, minimum one event).	$0.005

Event	What is charged	Free / no discount	Starter / Bronze	Scale / Silver	Business / Gold	Custom / Platinum	Custom / Diamond
item	Per Hugging Face model record produced	$0.575 / 1,000	$0.5 / 1,000	$0.39 / 1,000	$0.3 / 1,000	$0.2 / 1,000	$0.14 / 1,000

Apify may also charge platform usage for compute, storage, proxies, or data transfer outside this Actor pricing. Check the Actor run and the Apify Pricing tab for the exact cost shown to your account.

How to use Hugging Face Models Scraper

1. Open the actor on Apify.
2. Enter a search keyword such as llama, bert, or stable diffusion.
3. Optionally set an author such as google, microsoft, or meta-llama.
4. Optionally set a pipeline tag such as text-generation.
5. Choose a sort order.
6. Set a maximum number of models.
7. Run the actor.
8. Export the dataset or connect it to another system.

Input options

Search query

Use search to find models by names and tags. Examples:

• llama
• bert
• speech recognition
• stable diffusion
• embedding

Author / organization

Use author to focus on one Hugging Face namespace. Examples:

• google
• microsoft
• meta-llama
• stabilityai
• sentence-transformers

Pipeline tag

Use pipelineTag to filter by task. Examples:

• text-generation
• image-classification
• text-to-image
• automatic-speech-recognition
• sentence-similarity

Library / framework

Use library to filter by framework. Examples:

• transformers
• diffusers
• sentence-transformers
• timm
• pytorch

Sort order

Choose one of:

• downloads
• likes
• trending
• lastModified

Maximum models

Set limit to control output size. Start with 25 for a quick test. Use larger values for monitoring or market research exports.

Include detail data

Enable includeDetails when you need richer model-card metadata. This can add cardData and more complete file-count information. It may take longer because the actor checks individual model records.

Example input

{
  "search": "llama",
  "pipelineTag": "text-generation",
  "sort": "downloads",
  "limit": 25,
  "includeDetails": false
}

Example output

{
  "modelId": "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",
  "likes": 6214,
  "downloads": 9489535,
  "trendingScore": 24,
  "tags": ["transformers", "text-generation"],
  "pipelineTag": "text-generation",
  "libraryName": "transformers",
  "license": "llama3.1",
  "createdAt": "2024-07-18T08:56:00.000Z",
  "lastModified": null,
  "private": false,
  "gated": false,
  "sha": "...",
  "siblingsCount": 12
}

Tips for better results

• ✅ Use a specific keyword when you want focused results.
• ✅ Combine author and pipelineTag for organization-level monitoring.
• ✅ Sort by lastModified for update-monitoring workflows.
• ✅ Sort by downloads for popularity research.
• ✅ Turn on detail data only when you need richer metadata.

Common use cases

Model popularity tracking

Run the actor daily or weekly with the same query and compare downloads, likes, and trendingScore over time.

Competitive intelligence

Track models from specific organizations and monitor releases, tags, and licenses.

License monitoring

Export license and tags fields to find models that match or conflict with your internal compliance policies.

AI catalog enrichment

Use modelId, url, pipelineTag, and libraryName to enrich internal model catalogs or discovery products.

Integrations

Send results to Google Sheets, Slack, Make, Zapier, Airtable, BigQuery, Snowflake, or any system that accepts Apify dataset exports. Use scheduled runs for recurring snapshots. Use webhooks to trigger downstream processing when a scrape completes.

API usage with Node.js

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: process.env.APIFY_TOKEN });
const run = await client.actor('fetch_cat/hugging-face-models-scraper').call({
  search: 'llama',
  pipelineTag: 'text-generation',
  limit: 25
});
console.log(run.defaultDatasetId);

API usage with Python

from apify_client import ApifyClient
import os

client = ApifyClient(os.environ['APIFY_TOKEN'])
run = client.actor('fetch_cat/hugging-face-models-scraper').call(run_input={
    'search': 'llama',
    'pipelineTag': 'text-generation',
    'limit': 25,
})
print(run['defaultDatasetId'])

API usage with cURL

curl -X POST 'https://api.apify.com/v2/acts/fetch_cat~hugging-face-models-scraper/runs?token=YOUR_APIFY_TOKEN' \
  -H 'Content-Type: application/json' \
  -d '{"search":"llama","pipelineTag":"text-generation","limit":25}'

MCP usage

Connect Apify MCP to Claude Desktop or Claude Code and enable this actor as a tool. Use this MCP URL pattern:

https://mcp.apify.com/?tools=fetch_cat/hugging-face-models-scraper

Claude Code setup example:

$claude mcp add apify-hugging-face-models "https://mcp.apify.com/?tools=fetch_cat/hugging-face-models-scraper"

Claude Desktop JSON configuration example:

{
  "mcpServers": {
    "apify-hugging-face-models": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.apify.com/?tools=fetch_cat/hugging-face-models-scraper"
      ]
    }
  }
}

Example prompts:

• "Find the top downloaded Hugging Face text-generation models for llama."
• "Track the latest models from google on Hugging Face."
• "Create a CSV of popular sentence-transformers models with licenses."

Scheduling

Create an Apify schedule to run the same input daily, weekly, or monthly. Scheduled snapshots are useful for trend dashboards and model-change monitoring.

Data quality notes

The actor returns public metadata as reported by Hugging Face. Some fields may be missing for certain models. Gated models can still appear in public catalog data, but inaccessible private details are skipped.

Limits

The input limit is capped to keep runs predictable. Very broad searches can return many models, so choose a reasonable limit for your workflow. If you need very large recurring exports, start with a small test run and then increase gradually.

Troubleshooting

Why did I get fewer models than requested?

The selected filters may have fewer public results than your limit, or duplicate records may have been skipped. Try a broader query or remove one filter.

Why is cardData missing?

cardData is included only when Hugging Face returns it. Enable includeDetails for richer metadata, but remember detail runs take longer.

What if I see a rate-limit error?

Retry later or lower the limit. For frequent monitoring, schedule smaller runs rather than one very large run.

Legality

This actor collects publicly available metadata from Hugging Face. Make sure your usage complies with Hugging Face terms, Apify terms, and applicable laws. Do not use scraped data to violate model licenses or access restrictions.

Related Apify actors

Explore related Apify actors from the same catalog:

• https://apify.com/fetch_cat/github-repository-search-scraper
• https://apify.com/fetch_cat/arxiv-paper-search-scraper
• https://apify.com/fetch_cat/pubmed-search-scraper
• https://apify.com/fetch_cat/product-hunt-scraper

FAQ

Can I scrape private Hugging Face models?

No. This actor is designed for public model catalog metadata.

Can I filter by multiple authors at once?

Run the actor once per author or create multiple scheduled runs.

Can I export to CSV?

Yes. Apify datasets can be exported as CSV, JSON, Excel, XML, RSS, and HTML.

Can I monitor changes over time?

Yes. Schedule repeated runs and compare datasets by modelId.

Does this actor download model files?

No. It collects metadata only.

Is browser automation used?

No customer action is needed; the actor is optimized for reliable metadata collection.

Support

If a run fails or output looks wrong, open an issue with the run ID and the input you used. Include the expected result and a sample model URL if possible.