Public AI on Hugging Face Inference Providers 🔥

Source: HuggingFace Date: 2025-09-17 URL: https://huggingface.co/blog/inference-providers-publicai

Summary

Integration announcement: Public AI, a nonprofit inference provider backed by donated GPU time and institutional funding, joins HF Inference Providers. Currently free (rate-limited at 20 req/min with direct API key); HF credits apply. Hosts sovereign models from institutions like Swiss AI Initiative (Apertus-70B) and AI Singapore via vLLM on donated GPU clusters with global load balancing.

Implications

Thread: HF as open-source ML hub. Public AI as an inference provider is the clearest signal yet that the HF inference network is becoming a home for non-commercial and sovereign AI deployments, not just commercial cloud providers. Donated GPU + institutional funding is a fragile model long-term, but it creates a path for sovereign AI models to be usable without requiring cloud provider relationships. The Swiss AI and Singapore government models accessible through the standard InferenceClient are the most interesting examples — national AI investments becoming accessible to anyone via a HF token lowers the barrier to sovereign AI adoption significantly.