Hugging Face has officially integrated the Public AI project into its list of serverless inference providers. According to the development team—which includes Joseph Low, Joshua Tan, Julien Chaumond, and others—Hugging Face Hub users can now interact directly with sovereign AI models from organizations such as the Swiss AI Initiative and AI Singapore.
The Public AI Inference toolkit provides access to a global network of distributed computing powered by the vLLM library and OpenAI-compatible APIs. This creates a resilient deployment layer for models hosted on infrastructure provided by international institutional partners.
Public AI is a non-profit, open-source initiative. The system operates on GPU capacity allocated by government and industry institutions, supported by promotional subsidies. For technical professionals, this solution offers seamless integration via Python and JavaScript SDKs, utilizing global load balancing to automatically distribute requests across various compute clusters. Developers can manage the process directly through the Hugging Face interface, configuring their own provider API keys for direct calls or routing requests through their HF account.
This integration significantly streamlines access to public and sovereign models, eliminating infrastructure lock-in for developers. Public AI serves as a scalable bridge for implementing open models into production applications, ensuring transparent operation across distributed computing resources worldwide.