C.W.K.
Stream
Lesson 01 of 06 · published

What Hugging Face Actually Is

~30 min · hub, ecosystem

Level 0Scout
0 XP0/50 lessons0/10 achievements
0/120 XP to next level120 XP to go0% complete

The GitHub of AI, but with weights

Hugging Face (HF) is the open-weights operational hub of the modern AI stack. People reach for the cute analogy — it's GitHub for AI — but the platform is doing four jobs at once: repository host (Git-backed model/dataset/Space repos), discovery surface (search + filters across millions of repos), execution layer (inference widgets, Inference API, TGI, Spaces), and community platform (likes, discussions, Papers, Daily).

Conflating those four jobs is the most common entry mistake. People think "Hugging Face = pip install transformers" and miss everything that makes the platform sticky: the Hub is the registry, the moderation layer, the licensing surface, the leaderboard, and the abuse-control system. When you push a model card you are publishing a contract, not a README.

The numbers, calibrated

As of 2026 the Hub crossed 2 million public models, 500k+ datasets, 1M+ Spaces, and 13M+ registered users. About 30% of Fortune 500 companies have an org account. Numbers move fast — the relevant signal is "this is the central registry" not the exact count. Treat it like npm or PyPI: not a curated catalog, an open pipe with reputation gradients you read off cards, downloads, likes, and provider integration.

Every repo has Git history, branches, PRs, and diffs. That single design decision is why the Hub scales: tooling people already know, federated through a familiar mental model. The first thing to install is the Hub client — everything downstream rides on it.

Code

Install the Hub client and Transformers·bash
# CPU baseline — most lessons in this quest only need this
pip install --upgrade huggingface_hub transformers datasets

# Apple Silicon / MPS (skip on Linux)
pip install --upgrade torch torchvision torchaudio

# Login (interactive — pastes a token from huggingface.co/settings/tokens)
huggingface-cli login

# Confirm
huggingface-cli whoami
First contact: list trending models·python
from huggingface_hub import HfApi

api = HfApi()
trending = list(api.list_models(sort="trending", limit=10))
for m in trending:
    print(f"{m.id:<60} downloads={m.downloads or 0:>9}  likes={m.likes or 0:>5}")

External links

Exercise

Run huggingface-cli whoami after login. Then list the top 10 trending models, the top 10 trending datasets, and the top 10 trending Spaces. Save the three lists into a hf-trending-{date}.md file. Note: which categories surprise you, and which are dominated by a few orgs.

Progress

Progress is local-only — sign in to sync across devices.
Spotted a bug or have feedback on this page?Report an Issue

Comments 0

🔔 Reply notifications (sign in)
Sign inPlease sign in to comment.

No comments yet — be the first.