NaN

NaN is a paid community and shared GPU inference platform built for builders who want to run open AI models without the headache of managing their own infrastructure. Skip the server setup, capacity planning, and maintenance work, and start shipping instead.

Everything works through an OpenAI-compatible API, so you can plug it into your existing tools and workflows with minimal changes. Your requests run on shared dedicated GPUs, with all processing handled in the EU and zero prompt or response logging, keeping your data private by design.

Members get access to a broad model lineup covering language, embedding, reranking, text-to-speech, and speech-to-text models, so one subscription covers most of what a modern AI product needs. You can use the shared inference cluster whenever you need it.

Beyond the infrastructure, NaN is also a community. Join the private Discord to swap ideas with other builders, attend events, and vote on which models the platform adds next. For those who need more, a premium GLM 5.2 tier is available with published token allowances, so you always know exactly what you are getting.