Build without
limits.
Don't run out of tokens. Ever.
NaN is the community of builders sharing GPUs to run AI open models. Build whatever you need, with no token cap. You can just do things.
Limited spots Join the waitlist to unlock your subscription. Just leave your email.
01what is nan
What is NaN.
The Builders Community
Builders sharing dedicated GPUs to run open models and burn tokens without limits.
Closed-model APIs are convenient but expensive and limited. Open models are free but running them seriously requires hardware almost no one has at home.
NaN solves this by sharing powerful GPUs among builders who actually use them, splitting the real cost of infrastructure.
A closed community by design, limited to the real capacity of shared GPUs. When it fills up, we open a waitlist.
The community lives on Discord. The website only manages access. Every member pays to be here. Only those who truly want to build make it in.
02built with nan
We don't burn tokens.
We build.
Real projects built by NaN members using the shared inference cluster. This is what happens when builders get serious hardware.
NaN·MF2-HD1.44 MB Beyond The Protocol BeyondTheProtocol es una webapp médica privada construída para una paciente con cáncer metastásico ultra-raro.
Problema que resuelve: gestionar cientos de documentos clínicos dispersos (PDFs, análisis, reportes) de forma inteligente. Centraliza todo en un sistema RAG bilingüe (ES/EN) con búsqueda semántica, dashboard longitudinal, y un wiki clínico compilado, permitiendo explorar el historial médico completo con contexto y continuidad.
03the cluster
8 models.
One API.
The full open-model inference stack, voted by the community every quarter. Data stays in the EU, with nothing logged.
// llm
deepseek-v4-flash most used frontier
DeepSeek mimo-v2.5 most used frontier
Xiaomi gemma4
Google qwen3.6 most used
Alibaba // embedding
qwen3-embedding
Alibaba // rerank
qwen3-reranker
Alibaba // tts / stt
kokoro
Hexgrad whisper
OpenAI 04privacy
Pure inference.
Zero logs.
NaN is inference infrastructure. There's no logging of prompts, model responses, nothing. Everything lives in your local session, on whatever client, agent or custom app you use.
- Zero logs. No prompt or response logging.
- Everything lives in your local client or agent session.
- Processing in the European Union.
- Your code doesn't train any models.
- Server metrics only, for cluster maintenance (tokens/s, RPM).
05membership
One community.
Two tiers. One premium path.
The community is the product. Inference is the superpower you add; GLM 5.2 is the frontier option for members.
nan_member · glm 5.2 premium glm 5.2 · premium
200€ /mo · VAT included
- GLM 5.2, frontier open model with reasoning
- 3,000M token allowance per billing period
- 400M tokens per rolling 4h window: the limit a coding agent reaches first
- 500K context · 5 concurrent requests
- Everything in nan_member
- Access switches on a few minutes after payment
GLM 5.2 premium (200€/mo) is offered to members as seats open. Join the waitlist and we will flag you for this tier.
nan_member inference included
70€ /mo · VAT included
- Access to the shared cluster. Open models, no token caps.
- DeepSeek V4-Flash (frontier 284B, 1M context, reasoning). 500M tokens/month.
- Open models chosen by the community
- Personal API key compatible with OpenAI
- Private Discord channels for members only
- Events, workshops and hackathons
- Voice in the quarterly model vote
Month to month, no commitment. Billed in euros from any region. Limited spots: when it fills up, waitlist.
nan_community
14,99€ /mo · taxes included
- Discord access
- Member-only channels
- Live discussions, builds, AMAs
- Events, workshops and hackathons
- A reserved spot in line for inference
No commitment. Cancel anytime. Billed in euros from any region.
06who's behind
The NaN community is led by Cristian Córdova and Borja Perez.
I started NaN because I wanted to run serious open models without usage limits, and share and learn with people in the same boat: builders who actually burn tokens.
Cristian Córdova
Cristian Córdova
Founder of Helmcode
Founder of Helmcode, working with startups on cloud infrastructure, AI tooling and open source.
Thanks to Helmcode we offer open model inference with unlimited token plans for enterprises. I share my experiments on social media and AI research.
Borja Perez
Marketing Lead
Builder with marketing skills. Helping Cristian with comms and community.
Our goal is to create the best open-models community in the world, helping people create and ship products with open AI.
07faq
What you'll probably ask.
What is NaN?
A closed community of builders sharing dedicated GPUs to run open models. Every member pays to be here, and there's only room for people who actually build.
How does the cluster work?
The models run 24/7 on a shared inference cluster. Each member gets a personal API key compatible with OpenAI: change the base_url and you're in.
Is it really unlimited tokens?
On the cluster models, no counter. The frontier ones come with a token allowance per member so the plan holds up: 500M tokens per month on DeepSeek V4-Flash, 1.0B per month on MiMo v2.5, and 3,000M per billing period on GLM 5.2 for members on the premium tier. Those limits are published here, not in fine print. The bill is the fixed subscription price and it doesn’t move.
How are the models chosen?
Every quarter members vote on Discord which new models join the cluster. Candidates are pre-selected technically: they have to fit the real capacity. The open ecosystem moves fast, and so does the cluster.
Where does my data go?
Processed in the European Union, with no logging of prompts or responses. Your API key is personal and non-transferable. Your code trains no model and your data never leaves your session.
Who is NaN for?
Builders who want to run open models seriously without building the infrastructure: indie hackers, makers, people with agents and side projects in production. It's not for someone who just wants to try a couple of prompts.
Why only open models?
Because they can be audited, modified and used without asking permission. Closed ones are convenient but create dependency. With open models, the stack is yours.
Can I train models?
No. NaN is inference, not training. The GPUs are optimized to run existing models at high speed, not for fine-tuning.
What payment methods do you accept?
Credit or debit card. New memberships are billed in euros from any region, taxes included; subscriptions opened earlier in US dollars keep their original price. No commitment: cancel anytime.
How do I get in?
Leave your email on the waitlist and we'll write when a spot opens. There's a waitlist because there are no more GPUs: when compute comes in, people come in.
