Inference
DE

AI inference,
built for Europe.

German HQ 100% EU data residency

Open-weight AI models hosted on GPUs in the EU.

Zero data retention by default Chat completions are not stored. We process them in memory and discard them afterward. We never train on your data.
Create free account 100K tokens/month free
// models + pricing

Model catalog

We host open-weight models from Mistral, Google and IBM for reasoning, coding and multilingual work. The weights are open, the hosting is European, and your requests never leave the EU.

All models run on modern NVIDIA chips or newer. You pay per million tokens, and every model includes a free tier.

Mistral Google IBM
Dense
Mistral Medium 3.5 128B
Input: 1,10 € · Output: 6,10 €
Mistral Medium 3.5 128B
76
GPT-5 Mini
72
Sonnet 4.5
75
Gemini 3 Fl.
61
Mistral Medium 3.1
37
Mixture of Experts
Mistral Small 4 119B-6.5B
Input: 0,15 € · Output: 0,45 €
Mistral Small 4 119B-6.5B
66
GPT-5 Nano
47
Haiku 4.5
65
Gemini 3.1 FL
60
Mistral Small 3.2
26
Dense
Gemma 4 31B
Input: 0,15 € · Output: 0,35 €
Gemma 4 31B
74
GPT-5 Mini
72
Sonnet 4.5
75
Gemini 3 Fl.
61
Mistral Medium 3.1
37
Mixture of Experts
Gemma 4 26B-A4B
Input: 0,10 € · Output: 0,30 €
Gemma 4 26B-A4B
67
GPT-5 Nano
47
Haiku 4.5
65
Gemini 3.1 FL
60
Mistral Small 3.2
26
Dense
Granite 4.1 30B
Input: 0,25 € · Output: 1,30 €
Granite 4.1 30B
69
GPT-5 Mini
72
Haiku 4.5
65
Gemini 3 Fl.
61
Mistral Medium 3.1
37
Free tier
100K tokens/month All models 10 req/min No credit card

Prices in EUR, excl. VAT.

// for teams that need more

Business Plan

Fixed monthly pricing, reserved capacity when you need it, and a proper SLA. One plan covers all Nodion.ai products.

Business
500,00 € /month

Prices in EUR, excl. VAT.

Managed AI infrastructure for teams
  • 50M tokens/month included
  • Dedicated GPU capacity with priority routing
  • Model LTS: 12-month availability guarantee
  • Early access to new models
  • 99.5% uptime SLA
  • Dedicated support and onboarding
Get started
// getting started

API Documentation

Nodion.ai speaks the OpenAI API. Point your existing SDK or tool at our base URL and it should just work.

# Base URL
https://api.nodion.ai/v1
# Example: curl
curl https://api.nodion.ai/v1/chat/completions \
  -H "Authorization: Bearer $NODION_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "google/gemma4-31b",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Supports: /v1/chat/completions, /v1/models. Streaming, tool use, and JSON mode included.

// why this matters
Built for GDPR. We're a German company running GPUs in Germany, Sweden and Finland. Your data stays in the EU, full stop.
Nordic green energy. Our clusters run on renewable power. The cold climate helps with cooling, which keeps waste down.
No US dependency. European hosting, open-weight models, and no tie-in to a US hyperscaler.
Open weight only. We only host open-weight models. You know which checkpoint handles your requests, can reproduce the stack, and test the outputs yourself.
OpenAI-compatible API. Swap the base URL in your existing integration and you're on European infrastructure.

Ready to start?

100K free tokens per month. No credit card required. All models included.

Create free account