ETHUX
Sign inGet an API key
open models

Not your model? Not your knowledge.

Open models on our own GPUs in the Netherlands, GPU instances coming soon. One API key, pay per token.

Get an API key →See prices
Your prompts never leave the Netherlands.
Choose your AI

Open models, one API.

Pick a model, change one line in your code, done. The price per million tokens is next to each model.

ModelGood atPer million tokens
qwen3.8-27bGeneral chat and tool use. The default. Our own mixed quant, not FP8 or INT4.chattoolsreasoning256kmixed quant€ 0.10 in · € 1.20 out€ 0.05 cached in · excl. VATTry →qwen3-embeddingsVectors for search and retrieval.search32k€ 0.10 inexcl. VATTry →clef-flashA decision model. Give it a question and options, get a probability per option. You pay for input only.decisions9Binput onlycoming soon€ 0.04 inexcl. VATGet notified →
Sovereign AI

AI nobody can switch off for you.

Sovereign AI means you decide which model runs, where it runs and who can reach it. Not a vendor, not another government.

  • Open weights.The model file is public. Nobody can retire it or change it under you.
  • A Dutch company, a Dutch data centre.ETHUX is registered in the Netherlands, and the GPUs are here.
  • Or fully inside your walls.We set up the same stack on your own hardware.
Our own hardware

Bought, not rented.

The GPUs and servers are ours, in a Dutch data centre. So we know where your data is, and no cloud bill sets our prices.

0clouds in between
NLwhere every request runs
  • We own the GPUs.The GPUs and servers are ours, in a data centre in the Netherlands.
  • No cloud bill underneath.Our costs are hardware and power, not someone else's price list.
  • You can leave.Open models and a standard API, so your code works anywhere.
On your own hardware

Or run it on your own racks.

We deploy open models on your own GPUs: the model, the cluster underneath and the tool your team uses. Built by an NVIDIA-certified engineer.

On-prem deployments →
Everything ETHUX

What we build.

consoleConsole
Dashboard.Instances, models, keys and invoices in one place.
api.ethux.netAPI
OpenAI-compatible.One key for every model.
erebusErebus
The privacy filter.Use any model, keep personal data out.
docs.ethux.netDocs
How it works.Quick start, endpoints, limits.
Pricing

What it costs.

Excluding VAT. No minimums for the API; GPU instances will bill per started hour.

WhatUnitPrice
qwen3.8-27bper million tokens in / cached in / out€ 0.10 / € 0.05 / € 1.20
qwen3-embeddingsper million tokens€ 0.10
clef-flashper million tokens in, coming soon€ 0.04
1 × RTX 3090per hour, coming soon€ 0.35
2 × RTX 3090per hour, coming soon€ 0.70
Sandboxper hour, coming soon€ 0.02
Get an API key →