Hugging Face

Overview

Hugging Face is the hub of open-source AI. Its Inference API lets you call thousands of hosted models over HTTP — text generation, translation, summarization, image classification, zero-shot classification, object detection, speech-to-text, and more — without running anything yourself. The REST API also exposes model/dataset search and metadata.

Quick Start

Call a hosted model (replace YOUR_TOKEN; free token from huggingface.co/settings/tokens):

curl "https://api-inference.huggingface.co/models/gpt2" \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -d '{"inputs": "The meaning of life is"}'

Text classification (sentiment):

curl "https://api-inference.huggingface.co/models/distilbert-base-uncased-finetuned-sst-2-english" \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -d '{"inputs": "I love this movie!"}'

Image classification:

curl "https://api-inference.huggingface.co/models/google/vit-base-patch16-224" \
  -H "Authorization: Bearer YOUR_TOKEN" \
  --data-binary @cat.jpg

Search models on the hub:

curl "https://huggingface.co/api/models?search=text-generation&limit=5"

Authentication

• Create a free account and generate a token (settings → Access Tokens).
• Send it as Authorization: Bearer <TOKEN>.
• The free tier includes a monthly inference quota (exact amounts on the pricing page); some models are gated.

Rate Limits & Notes

• Free tier: limited monthly inference quota; paid tiers add capacity and faster models.
• Inference is serverless — models may cold-start on first call (seconds).
• You can also run models yourself (transformers / TGI) and point the same client code at your endpoint.
• Model availability changes — check the model page for gating and task support.

FAQ

Is Hugging Face free? Yes — free account with a monthly inference quota; paid tiers for more.

Which models can I call? Thousands of hosted models — text, vision, audio. Search the hub (/api/models?search=...) or the website.

Can I self-host instead? Yes — the same models run locally with the transformers library or a TGI server.