Pricing — per million tokens, no subscription | europeanGPU

One price per token. That's it.

No subscription, no credits to preload, no exit cost. You are billed in euros, every month, for the tokens you actually consume — and you can leave whenever you want.

ModelInput / M tokensOutput / M tokensContext
Qwen3.8-27B
Reasoning · code · 100+ languages
0,38 € 2,60 € 262 K Model card →
Model sovereignty →
Ministral 3 14B Instruct 2512
European publisher · fast and frugal · 128 K context
0,20 € 1,30 € 128 K Model card →
Model sovereignty →
More open models will join the catalogue — tell us which ones during your meeting.

Prices excl. VAT, monthly billing on actual usage, per-token granularity.

Three concrete reference points

~0,94 €

for 1 M tokens
of a typical chatbot

~2 600

A4 pages summarised
for 1 €

0 €

in fixed, entry
or exit fees

Frequently asked questions

How are tokens counted?

With the model's tokenizer, identical to the one you can run locally. Every API response returns the exact count consumed.

Are there volume tiers?

Beyond 5 billion tokens per month, negotiated pricing and reserved capacity are possible — that is what the sales meeting is for.

Does reasoning mode cost more?

No: same per-token prices. Reasoning tokens are simply counted as output tokens.

Where is my data processed?

Exclusively in our European Union datacenters, by an entity incorporated under European law. The details here.

Large volumes or specific constraints?

Book a meeting: together we scope volumes, latency and compliance, and open a trial access.

Book a meeting