Inference that stays in Europe

Run open-source models on a sovereign European cloud through one API. Start on shared endpoints, move to dedicated GPUs when you scale.

Run your first modelTalk to us
Served in EuropePrompts and answers stay inside the EU.
Proven on every requestEach request records the country that served it.
Works with your codeSame format as OpenAI. Change the URL and the key.
Pay for what you usePrepaid credits. No minimum term.
Runs the open models you already know
and more

What you get from open models in Europe

Four reasons teams move their AI workloads closer to home.

01
Your data stays putProcessed and logged in the EU. Choose Germany, France or Finland.
02
Lower costOpen models cost less to run. We're built to close the EU/US price gap.
03
Faster answersModels run close to European users. Watch the first word arrive.
04
Yours to shapeFine-tune an open model on your own data and GPUs.

Start small. Own it when you're ready.

Most teams begin on Model endpoints and move to their own GPUs once traffic is steady. Same model, same API, no migration.

Billed per tokenShared with other requests
Model endpoints

Shared endpoints for open-source models

Call popular models straight away. Nothing to provision, nothing idle.

  • Kimi, DeepSeek and GLM ready to call
  • Pay for input and output tokens, nothing else
  • Test prompts in the Playground before you write code
KimiDeepSeekGLM
Run your first model
Billed per GPU hourReserved for you
Dedicated

Your own GPUs, in the region you pick

Choose a model, choose a European region. The GPU comes set for that configuration, so you skip the tuning.

  • Pre-optimised templates for the models we serve
  • Capacity that is yours, with no noisy neighbours
  • Bring your own Hugging Face model Soon
H100H200B200
Talk to us
01 → PlaygroundTry it on your own prompts
02 → Model endpointsCreate a key, call a shared model
03 → DedicatedMove the same model to its own GPUs

Every request, logged where it ran

Each line shows the model, the key and the country that served it. Anything processed outside the EU gets flagged.

100%of requests carry a processing location
ModelProcessed inStatus
kimi-k2GermanyIn EU
deepseek-v3FranceIn EU
glm-4.6FinlandIn EU
kimi-k2GermanyIn EU
kimi-k2Outside EUFlagged
deepseek-v3FranceIn EU
glm-4.6FinlandIn EU
kimi-k2GermanyIn EU
deepseek-v3FranceIn EU
glm-4.6GermanyIn EU
kimi-k2FinlandIn EU
deepseek-v3Outside EUFlagged
glm-4.6FranceIn EU
kimi-k2GermanyIn EU
deepseek-v3FinlandIn EU
glm-4.6FranceIn EU
kimi-k2GermanyIn EU
deepseek-v3FranceIn EU
glm-4.6FinlandIn EU
kimi-k2GermanyIn EU
kimi-k2Outside EUFlagged
deepseek-v3FranceIn EU
glm-4.6FinlandIn EU
kimi-k2GermanyIn EU
deepseek-v3FranceIn EU
glm-4.6GermanyIn EU
kimi-k2FinlandIn EU
deepseek-v3Outside EUFlagged
glm-4.6FranceIn EU
kimi-k2GermanyIn EU
deepseek-v3FinlandIn EU
glm-4.6FranceIn EU

Evidence for GDPR records of processing and EU AI Act reviews, without building your own tracing.

Tell Hanse what matters. It picks the setup.

Give each workload a priority and keep one endpoint.

01Default
Cost
The lowest price per request.For batch jobs, evals and background agents.
02
Performance
The shortest time to first token.For chat, voice and anything a person waits on.
03
Quality
The most faithful model setup.For reasoning, tool calls and long context.

Simple prices. Europe shouldn't cost more.

Running AI in Europe has cost about 30% more than in the US. We're closing that gap.

Model endpoints
from €0.01 / 1M tokens
Input and output tokens, nothing else
Dedicated
Custom pricingContact sales →
Billed per second while you reserve it
Popular modelsEUR per 1M tokens · excl. VAT · billed per token
DeepSeek V4 FlashDeepSeekInput€0.10Cached€0.02Output€0.25
Qwen3.8 27BQwenInput€0.33Cached€0.081Output€1.94
GLM-5.3Z.aiInput€1.00Cached€0.13Output€3.00
Kimi K3.1MoonshotInput€2.42Cached€0.61Output€12.08
See all prices

Questions, answered

Move your AI to Europe.

One key, one endpoint, three European cities. Your first call takes minutes.

Get an API keyTalk to us