Over 100 Million Developers on GitHub Can Access Top Global Models to Build AI Applications
"GitHub Models" accelerates AI engineering with top industry models, easy integration, and seamless deployment via Azure. Join the AI revolution!
GitHub's new feature, "GitHub Models," is set to speed up the AI engineer era.
Wait, what? The familiar code hosting platform GitHub has evolved! Now it offers an AI model playground.
All the popular industry models you know, like Microsoft's Phi-3, OpenAI's GPT-4o, Meta's Llama 3.1, Cohere's Command R+, and Mistral AI's Mistral Large, are available for testing in an interactive sandbox.
In the coming months, GitHub will add more language, vision, and other types of models.
This means all these models are now free to use! It's a new way to test various models for free.
Moreover, developers can easily import the right model into their AI projects.
GitHub has created a fast track to place models directly into Codespaces and VS Code development environments, lowering the barrier to deploying AI models.
This is the new "GitHub Models" feature launched by GitHub today!
With the right tools and training, any developer can become an AI engineer.
Developers can choose a model, code in GitHub Codespaces, and deploy to production via Azure, offering an end-to-end service.
Specifically, you can use the GitHub CLI to call GitHub Models commands, use them in GitHub Actions through a series of JSON files, and build GitHub Copilot extensions, covering the entire app development lifecycle.
When the project is ready for launch, GitHub provides seamless integration with Azure AI. GitHub authentication via an Azure subscription makes AI application deployment to production easier.
Now, you can deploy AI applications in over 25 Azure regions worldwide through GitHub and get Azure's enterprise-level security.
In line with GitHub and Microsoft's ongoing commitment to privacy and security, no prompts or outputs from GitHub Models will be shared with model providers or used to train or improve models.
Of course, using "GitHub Models" has some limits. Individual users are limited to 150 accesses per day, 15 times per minute, with each request handling up to 8000 tokens and outputting up to 4000 tokens.
GitHub CEO Thomas Dohmke said this feature marks another transformation for GitHub, from creating AI through open-source collaboration to using AI to create software, and now to fostering AI engineers with GitHub Models.
GitHub aims to attract as many partners as possible to its platform based on OSS repositories, Copilot Extensions, and GitHub Models.
Some developers have shown examples of running models directly in GitHub Codespaces without installation, starting in seconds.
How will Hugging Face respond to GitHub's new move? We'll have to wait and see.