You built a cool ML demo. Now you need a server, a domain, SSL, and a Dockerfile. Or you could spend 3 minutes on Hugging Face Spaces and get a public URL with free CPU (or GPU if you're lucky).
The 3-Minute Deploy
pip install gradio huggingface_hub
huggingface-cli login # token from hf.co/settings/tokens
# app.py
import gradio as gr
from transformers import pipeline
classifier = pipeline("sentiment-analysis", model="distilbert-base-uncased-finetuned-sst-2-english")
def predict(text):
return classifier(text)[0]
gr.Interface(fn=predict, inputs="text", outputs="label", title="Sentiment Demo").launch()
huggingface-cli repo create my-demo --type space
git clone https://huggingface.co/spaces/YOUR_USERNAME/my-demo
cd my-demo
# copy app.py + requirements.txt
git add . && git commit -m "init" && git push
Your demo is live at https://YOUR_USERNAME-my-demo.hf.space in ~60 seconds.
What You Get for Free
| Feature | Free Tier |
|---|---|
| CPU | 2 vCPU, 16GB RAM |
| GPU | T4 (community grant, queue-based) |
| Custom domain | Yes (CNAME) |
| Sleep after | 48h inactivity |
| Persistent storage | 50GB (ephemeral) |
The Catch
The free T4 GPU is a lottery — you request a community grant and wait. For CPU-only demos (sentiment, classification, small LLMs with llama.cpp) it's perfect. For SDXL or 70B models, bring your own inference API.
I use Spaces for every side project demo now. Zero DevOps, zero cost, and the URL looks legit in a portfolio.
Built the demo UI with MonkeyCode: https://ly.cyberserval.tech/iIETXiF
What's the most creative thing you've seen hosted on Spaces?
Top comments (0)