Become a Model Partner

Launching models with the Public AI Inference Utility

General

What is this about?

The Public AI Inference Utility provides stable, public access to AI models built by national labs, public research institutions, and governments. By collaborating with us, your model can be deployed globally in a way that is technically robust, transparent, and aligned with public-interest values.

Who are typical model partners?

National supercomputing centers, government-funded AI institutes, universities, and government ministries or agencies commissioning model development. Our current partners include Switzerland (Swiss National AI Institute) and Singapore (AI Singapore).

We prioritize partners that bring:

  • Impact on culture – Models with a clear national or scientific identity.
  • A news hook or tie-in – Launches aligned with national events, AI strategies, or regional initiatives.
  • Government support – Endorsement or sponsorship from public institutions, ensuring legitimacy and long-term sustainability.

Why partner with Public AI?

Because deploying a model at scale requires more than training it. The Public AI Inference Utility:

  • Handles user-facing infrastructure, including frontend, routing, failover, data storage, and compliance with relevant laws such as GDPR.
  • Provides stable APIs (OpenAI-compatible) and a tested deployment stack.
  • Integrates automatically with Hugging Face through our status as an official inference provider.
  • Supports launch activities (PR, media, community engagement).
  • Manages inference partners (cloud providers and HPC centers contributing compute).
  • Ensures your model is visible, discoverable, and used.

"So many people have been using Public AI to get an idea of the model. It's been extremely valuable to just get the word out and get people to play around with it. People are already building on the API. There's actually a very nice use-case that came out immediately: a company building on open-source models has a system for answering tax questions in Zurich. That's now our go-to use case, because it shows links, hallucinates very little, and highlights what you can build with [Apertus]. I was actually surprised it worked so well out of the box."

Imanol Schlag— Imanol Schlag, co-lead of Apertus, Swiss AI Initiative

Do we need to provide compute?

Not necessarily. Public AI partners with inference providers worldwide who donate compute. However, to ensure a successful pilot period with free compute, we highly encourage working with us to rally inference partners to support your launch.

If your institution has its own compute that it wishes to donate (or funds to purchase compute), we can add it directly to the routing pool.

Next Steps

How do we become a Model Partner?

  • To express interest, contact: josh@publicai.co.