HuggingFace Intelligence Scraper — Prime Sieve

Scrape HuggingFace Models, Datasets, Spaces, and Trending AI repos — no login, no API key, no browser.

Built by Prime Sieve. Uses HuggingFace's public /api/models|datasets|spaces|trending endpoints (fetch-clean). Ideal for the AI/dev ecosystem: model leaderboards, dataset research, and finding what's trending.

What you get

Choose a mode and get structured rows per record:

models — id, author, pipeline tag, library, tags, gated/private, likes, downloads, trending score, parameter count, created/lastModified, URL. datasets — id, author, pipeline tag, tags, gated/private, likes, downloads, trending score, description, lastModified, URL. spaces — id, author, SDK (e.g. docker/streamlit), tags, likes, trending score, URL. trending — what's actively trending on HuggingFace right now: repo type, ID, author, likes, downloads, param count, isLikedByUser, URL.

Pricing

$0.0015 per result — pay per result, no subscriptions.

	Prime Sieve (this)	david_flagg/huggingface-models
Price	$0.0015 / result	$0.001 / result
Models	✅	✅
Datasets	✅	❌
Spaces	✅	❌
Trending	✅	❌
Sort (downloads/likes/trending)	✅	❌
Cursor pagination to 1,000	✅	❌
Maintenance	active	stale

Ships the full HuggingFace surface (all 4 modes + sorting) at a comparable unit price — more value per run, one actor instead of several.

Use cases

1. LLM watch / leaderboard — track top models by downloads or trending score, monitor new releases
2. Dataset research — find candidate datasets by keyword, sort by downloads/likes for what's trusted
3. AI startup intel — see which frontier orgs are releasing models/spaces, parameter counts, gating
4. Competitive monitoring — scheduled runs diff likes/downloads/trending over time

Input

{
  "mode": "models",
  "search": "llama",
  "limit": 100,
  "sortBy": "downloads",
  "direction": -1,
  "full": true,
  "timeoutSeconds": 20
}

Field	Default	Notes
mode	models	models / datasets / spaces / trending
search	``	keyword filter
limit	20	max 1,000
sortBy	``	downloads / likes / trendingScore (or createdAt for spaces)
direction	``	-1 desc, 1 asc
full	true	include description + full tags
timeoutSeconds	20	per-request timeout

Output sample (models)

{
  "mode": "models",
  "id": "meta-llama/Llama-3.1-8B-Instruct",
  "author": "meta-llama",
  "pipelineTag": "text-generation",
  "libraryName": "transformers",
  "tags": ["transformers", "safetensors", "llama", "text-generation"],
  "gated": "manual",
  "private": false,
  "likes": 6848,
  "downloads": 5644039,
  "trendingScore": 41,
  "numParameters": null,
  "createdAt": "2024-07-18T08:56:00.000Z",
  "lastModified": "2024-09-25T17:00:57.000Z",
  "url": "https://huggingface.co/meta-llama/Llama-3.1-8B-Instruct"
}

---

📬 Get Remote Tech Roles & Hiring Signal Weekly

Curated remote engineering roles, tech stacks, and market data delivered every Wednesday. No fluff. 👉 Subscribe to Remote Signal