Inference Providers

Hugging Face Inference Providers simplify and unify how developers access and run machine learning models by offering a unified, flexible interface to multiple serverless inference providers. This new approach extends our previous Serverless Inference API, providing more models, increased performances and better reliability thanks to our inference partners.

To learn more about the launch of Inference Providers, check out our announcement blog post.

Why use Inference Providers?

Inference Providers offers a fast and simple way to explore thousands of models for a variety of tasks. Whether you’re experimenting with ML capabilities or building a new application, this API gives you instant access to high-performing models across multiple domains:

⚡ Fast and Free to Get Started: Inference Providers comes with a free-tier and additional included credits for PRO users, as well as Enterprise Hub organizations.

Key Features

Inference Playground

To get started quickly with Chat Completion models, use the Inference Playground to easily test and compare models with your prompts.

Get Started

You can use Inference Providers with your preferred tools, such as Python, JavaScript, or cURL. To simplify integration, we offer both a Python SDK (huggingface_hub) and a JavaScript SDK (huggingface.js).

In this section, we will demonstrate a simple example using deepseek-ai/DeepSeek-V3-0324, a conversational Large Language Model. For the example, we will use Novita AI as Inference Provider.

Authentication

Inference Providers requires passing a user token in the request headers. You can generate a token by signing up on the Hugging Face website and going to the settings page. We recommend creating a fine-grained token with the scope to Make calls to Inference Providers.

For more details about user tokens, check out this guide.

cURL

Let’s start with a cURL command highlighting the raw HTTP request. You can adapt this request to be run with the tool of your choice.

curl https://router.huggingface.co/novita/v3/openai/chat/completions \
    -H "Authorization: Bearer $HF_TOKEN" \
    -H 'Content-Type: application/json' \
    -d '{
        "messages": [
            {
                "role": "user",
                "content": "How many G in huggingface?"
            }
        ],
        "model": "deepseek/deepseek-v3-0324",
        "stream": false
    }'

Python

In Python, you can use the requests library to make raw requests to the API:

import requests

API_URL = "https://router.huggingface.co/novita/v3/openai/chat/completions"
headers = {"Authorization": "Bearer hf_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"}
payload = {
    "messages": [
        {
            "role": "user",
            "content": "How many 'G's in 'huggingface'?"
        }
    ],
    "model": "deepseek/deepseek-v3-0324",
}

response = requests.post(API_URL, headers=headers, json=payload)
print(response.json()["choices"][0]["message"])

For convenience, the Python library huggingface_hub provides an InferenceClient that handles inference for you. Make sure to install it with pip install huggingface_hub.

from huggingface_hub import InferenceClient

client = InferenceClient(
    provider="novita",
    api_key="hf_xxxxxxxxxxxxxxxxxxxxxxxx",
)

completion = client.chat.completions.create(
    model="deepseek-ai/DeepSeek-V3-0324",
    messages=[
        {
            "role": "user",
            "content": "How many 'G's in 'huggingface'?"
        }
    ],
)

print(completion.choices[0].message)

JavaScript

In JS, you can use the fetch library to make raw requests to the API:

import fetch from "node-fetch";

const response = await fetch(
    "https://router.huggingface.co/novita/v3/openai/chat/completions",
    {
        method: "POST",
        headers: {
            Authorization: `Bearer hf_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx`,
            "Content-Type": "application/json",
        },
        body: JSON.stringify({
            provider: "novita",
            model: "deepseek-ai/DeepSeek-V3-0324",
            messages: [
                {
                    role: "user",
                    content: "How many 'G's in 'huggingface'?",
                },
            ],
        }),
    }
);
console.log(await response.json());

For convenience, the JS library @huggingface/inference provides an InferenceClient that handles inference for you. You can install it with npm install @huggingface/inference.

import { InferenceClient } from "@huggingface/inference";

const client = new InferenceClient("hf_xxxxxxxxxxxxxxxxxxxxxxxx");

const chatCompletion = await client.chatCompletion({
    provider: "novita",
    model: "deepseek-ai/DeepSeek-V3-0324",
    messages: [
        {
            role: "user",
            content: "How many 'G's in 'huggingface'?",
        },
    ],
});

console.log(chatCompletion.choices[0].message);

Next Steps

In this introduction, we’ve covered the basics of Inference Providers. To learn more about this service, check out our guides and API Reference:

< > Update on GitHub