DeepSeek-R1 and GitHub Models: A New Alternative to Hugging Face?

GitHub has entered the AI model hosting space with GitHub Models, a platform that allows developers to find, experiment with, and integrate AI models directly into their workflows. The latest addition to this service is DeepSeek-R1, a powerful large language model (LLM) now available in public preview.

What is GitHub Models?

If you're familiar with Hugging Face (with its vast collection of AI models, more than 1.3 million models as when this article was written), then you'll immediately recognize the intent behind GitHub Models. This service enables developers to:

• Find AI models for various tasks (NLP, code generation, etc.).
• Try, compare, and implement models via a playground or API.
• Use AI models from the command line with GitHub CLI or integrate them as Copilot extensions.
• Leverage a free API (currently in public preview) to test AI capabilities.

Essentially, it's Microsoft/GitHub's answer to Hugging Face—a move that could shift where developers go for AI integrations. Although GitHub Models is still in its early stages, the addition of DeepSeek-R1 is a significant step forward. At the time of writing this article, I counted only 22 models available in the GitHub Models' marketplace.

Why DeepSeek-R1 Matters 🐋

DeepSeek-R1 is an advanced large-scale AI model designed to push the boundaries of deep learning, NLP, and computer vision. Here's why it's important:

1. Massive Scale – This model is among the most potent publicly available LLMs.
2. Free Public Preview – Developers can experiment with it for free via GitHub Models' playground or API.
3. Side-by-Side Comparisons – GitHub Models allows users to compare DeepSeek-R1 against other models.
4. Command-Line and Copilot Integration – You can test and implement it directly using GitHub CLI or within GitHub Copilot.
5. Future of AI Model Hosting – This release signals GitHub's ambition to compete with Hugging Face and become the go-to hub for AI model distribution.

GitHub Models vs. Hugging Face: The Bigger Picture

Microsoft's AI Strategy

With GitHub Models, Microsoft is expanding its AI ecosystem, making it easier for developers to consume LLMs without leaving GitHub. This aligns with their broader strategy:

• Deep OpenAI Partnership – Microsoft has invested heavily in OpenAI and is integrating AI models directly into its products.
• GitHub Copilot – AI-assisted coding is already integrated into GitHub; now, models like DeepSeek-R1 will enhance this further.
• Azure AI Integration? – It's likely that GitHub Models will eventually tie into Azure AI, pushing enterprise adoption.

Hugging Face's Edge

Hugging Face has been the dominant developer-first AI model hub. It has:

• A strong open-source community.
• A large collection of pre-trained models.
• Collaborative tools for fine-tuning and deploying models.

The key battle will be ecosystem lock-in - GitHub has a vast user base, but Hugging Face has deep community trust.

Getting Started with DeepSeek-R1 on GitHub Models

Want to try DeepSeek-R1 yourself? Here's how:

• Use the Playground – Experiment with the model for free.
• Try the API – Integrate DeepSeek-R1 into your own applications.
• Compare AI Models – Test its performance against other models with GitHub's side-by-side comparison feature.
• Run from CLI – Use the GitHub CLI for quick testing.
• Extend Copilot – Integrate DeepSeek-R1 as a Copilot extension.

👉 Official GitHub Documentation

I tested the chat section of DeepSeek-R1 with the following prompt:

tip

"Can you explain Agentic AI and how I can start integrating it with Agentico.dev (documentation: https://agentico.dev/docs)?"

It was a way for me to validate the model's understanding of a complex question, and to test if my documentation is clear an enough. 😅

The response was impressive, showing the model's ability to comprehend and generate human-like text, did you notice it's reasoning about the documentation and how it's line of thought is structured? 🤯

Well, the bad news is that it can't access the internet, so it can't check the documentation for you. 🥺

👨🏻‍💻 Sample Code: Running DeepSeek-R1 in JavaScript

To authenticate with the model you will need to generate a personal access token (PAT) in your GitHub settings. Create your PAT token by following instructions here: Creating a personal access token.

This is the example you can find in the GitHub Models' playground:

/*
$ npm install @azure-rest/ai-inference @azure/core-auth @azure/core-sse
*/
import ModelClient from "@azure-rest/ai-inference";
import { AzureKeyCredential } from "@azure/core-auth";
// Your PAT token
const token = process.env["GITHUB_TOKEN"];

export async function main() {
    const client = new ModelClient(
        "https://models.inference.ai.azure.com",
        new AzureKeyCredential(token)
    );

    const response = await client.path("/chat/completions").post({
        body: {
            messages: [
                { role: "user", content: "Can you explain what Agentic AI is and how I can start integrating it with Agentico.dev (documentation: https://agentico.dev/docs?)" }
            ],
            model: "DeepSeek-R1",
            max_tokens: 2048,
        }
    });

    if (response.status !== "200") {
        throw response.body.error;
    }
    console.log(response.body.choices[0].message.content);
}

main().catch((err) => {
    console.error("The sample encountered an error:", err);
});

Final Thoughts: A New AI Hub in the Making?

GitHub Models is a significant move in the AI model space. It gives developers a new way to discover, compare, and integrate AI models - right where they already code. While Hugging Face remains the leader in open-source AI, GitHub Models has the potential to disrupt the landscape, especially with Microsoft's backing and deep integration with Copilot and GitHub Actions.

Will GitHub Models become the preferred AI hub for developers? Time will tell, but the competition is heating up. 🔥

Have you tried DeepSeek-R1 yet? Let us know your thoughts! 🚀