Isolated resources
A GPU reserved for your workload, with consistent performance for your traffic.
Project-specific hosting scope
Your own GPU for private AI in Europe
For teams that need isolated resources: a fully managed OpenAI-compatible endpoint on a GPU reserved for you, with hosting location agreed with you and a DPA.
- Operated by a German company
- GDPR · DPA available
- ISO 27001-certified data centres
- Hosting location agreed with you
Availability terms
Service levels are agreed per project and written into your contract.
German company, DPA
Personal data is processed under the GDPR. A DPA with TOMs is available.
Unlimited Tokens
Fixed hourly rate, no per-token charges. No rate limits.
Where your endpoint runs
You know where your endpoint runs.
Falkenstein
Germany · Saxony
- GDPR Article 28 DPA available
- Close to DACH & Eastern Europe
- ISO 27001-certified data centre; LLMBase works to the standard
Helsinki
Finland · Northern Europe
- GDPR Article 28 DPA available
- Close to Nordics & Baltic states
How it works
A dedicated endpoint shaped around your requirements.
01
Define your endpoint
Tell us which open model you need, how it should perform, and which access controls matter to your team.
02
Agree your hosting location
Hosting location agreed with you, based on your data-residency and latency requirements. Your resources stay isolated for your workload.
03
Go live with one API
We configure and operate the endpoint. Your team receives a standard OpenAI-compatible API without infrastructure work.
Dedicated Inference
When to choose Dedicated
A fully managed endpoint on a GPU reserved exclusively for you. LLMBase handles deployment, model loading, and operations — you get a standard OpenAI-compatible API. No SSH access, no container management.
- Steady, high-throughput workloads running continuously
- Consistent, predictable latency on every request
- Your own fine-tuned or custom model weights
- Full resource isolation for compliance or security
- Fixed hourly cost, not per-token billing
Serverless Inference
When to choose Serverless
Send requests to shared GPU infrastructure. No setup required — get an API key and start in minutes. You only pay for the tokens you generate.
- Getting started quickly without any infrastructure setup
- Unpredictable or spiky traffic patterns
- Low-volume, experimental, or batch workloads
- Foundation models only — no custom weights needed
- Paying only for tokens consumed