Hugging Face Inference

Provider

One API for open models served by many inference providers

Price: Free tier, then paid
Access: User access token

About

Hugging Face's serverless API that routes requests for open-weight models to partner providers such as Groq, Together and fal, with one token and one bill. Suits developers who want many open models without separate accounts. The monthly free credit is small.

What you can do with it

Call open-weight LLMs like gpt-oss and DeepSeek through one OpenAI-compatible endpoint
Pick the fastest or cheapest provider per model with a suffix like :cheapest
Generate images, embeddings and transcripts from Hub models in Python or JS

Get started

Create a fine-grained token with Inference Providers permission
Pick a model that a provider serves on the Hub

Example

curl https://router.huggingface.co/v1/chat/completions \
  -H "Authorization: Bearer $HF_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-oss-120b:fastest",
    "messages": [{"role": "user", "content": "How many G in huggingface?"}],
    "stream": false
  }'

Details

Hosting: Hosted service
Available in: Worldwide
Official SDKs: Python, JavaScript/TypeScript
MCP server: None