You need GPU power for your own language model, but not a server your team has to look after at night. We provide dedicated NVIDIA GPU servers in a German data centre, with a tuned model, vLLM, Open WebUI, 24/7 monitoring and updates.
Configurations and pricing
Both configurations include the dedicated GPU server, one agreed model, vLLM, Open WebUI and managed operations by WZ-IT.
A compact starting point for a dedicated, fully managed model and inference environment.
A managed model and inference environment for larger models, longer contexts and higher concurrency.
All prices are net and exclude statutory VAT. The offers are addressed to businesses.
Multi-GPU servers
Dedicated servers with two, three or four NVIDIA RTX PRO 6000 Blackwell Max-Q in a German data centre, with vLLM, Open WebUI and managed operations by WZ-IT. CPU, memory and storage are configured to match the workload; price, setup and term are stated in the proposal. Field report: Qwen3.5-122B on two RTX PRO 6000
One large model across both GPUs via tensor parallelism, for example a 120B-class model in FP8 with headroom for context and concurrent users.
192 GB GPU memoryPrice on requestRequest in the configurator 3 × RTX PRO 6000 Blackwell Max-QSeveral models side by side, each on its own GPU, for example a chat, a vision and an embedding model. For a single large model, two or four GPUs are the right choice.
288 GB GPU memoryPrice on requestRequest in the configurator 4 × RTX PRO 6000 Blackwell Max-QVery large models, or one large model with many concurrent users and long contexts, via tensor parallelism across all four GPUs.
384 GB GPU memoryPrice on requestRequest in the configuratorManaged operations
WZ-IT does not only provide compute capacity. We keep the agreed technical foundation traceable and supportable throughout ongoing operations.
Both configurations are provisioned as usable, managed model environments. Model selection remains dependent on GPU memory and the actual workload.
Model and vLLMOne agreed model is delivered through a managed vLLM inference layer. API, runtime, updates and monitoring are included.
View managed vLLMView complete LLM hosting
Open WebUIA managed chat and user interface for the deployed model. Baseline configuration, operations, updates and monitoring are included.
More about Open WebUILinux, administrative access, baseline firewall protection, NVIDIA drivers, vLLM, Open WebUI and one agreed model are configured.
The server is located in Germany. External APIs, telemetry, integrations and backup targets are treated and agreed as separate data paths.
Host, GPU health, capacity and agreed services are monitored proactively. Relevant events enter the operating process.
Incidents are accepted, assessed and handled with documentation according to the agreed service level.
The operating system, drivers, vLLM and Open WebUI are checked against security advisories and updated in a controlled manner. Critical application vulnerabilities are prioritised.
Configuration, responsibilities and relevant changes remain traceable. Your point of contact knows the agreed environment.
Setup
Setup
The one-time setup creates a documented and monitored model environment. Custom integrations are scoped separately.
Check model, quantisation, context, concurrency and additional compute loads against 24 or 96 GB of GPU memory.
Configure the host system, Linux, access, network, NVIDIA drivers, vLLM, Open WebUI and the agreed model.
Set up monitoring, alert paths, maintenance windows, service level and operations documentation.
Jointly verify the GPU, model endpoint, Open WebUI, reachability and monitoring, then hand over the agreed state.
Use case and sizing
Parameter count alone is not enough. Quantisation, context length, concurrent users, batch size, KV cache and runtime determine memory and performance requirements.
Model, quantisation, context and concurrency in detail:
Explore VRAM sizing in the knowledge baseOperating model
A fixed managed GPU server is not the most economical form for every workload. Utilisation and required operational responsibility determine the right model.
A rented GPU server is only part of the cost. Running it yourself adds working time and on-call duty that appear in no price list.
| Rent a server and run it yourself | Managed GPU Server |
|---|---|
| Monthly server rent at the hosting provider | One monthly price for server and operations, plus a one-time setup |
| Setup of operating system, drivers, CUDA, vLLM, Open WebUI, TLS and access by your team | Model, vLLM and Open WebUI set up and tuned |
| Ongoing driver, vLLM and security updates with your own CVE assessment | Updates of operating system, drivers, vLLM and Open WebUI with CVE assessment |
| Monitoring and on-call duty, including nights and weekends | Proactive 24/7 monitoring and incident handling |
| Model changes, testing and documentation, often depending on one person | Documentation and a named contact |
Classify the service
Each page has a distinct role, keeping hardware, model serving and ongoing platform operations out of one ambiguous offer.
When you need an operated model endpoint, an OpenAI-compatible API or a complete inference layer.
View serviceWhen an existing or individually designed AI platform with models, runtimes and applications should be taken into operations.
View serviceWhen the AI hardware should remain at your site and operate as a local appliance.
View servicePricing, scope, GPU selection, access and extensions
Knowledge base
Reviews & projects
From a knowledge base with an AI assistant and local AI integration to data sovereignty and ongoing operations.
AI Platform
We were looking for a secure yet easy-to-use alternative to the large American providers and came across WZ-IT. WZ-IT advised us on choosing the right models and implemented the integration for us. Beforehand, we were able to try everything out calmly on a test instance. Today WZ-IT runs Open WebUI for us on our own domain, with a convenient Microsoft login and open-source models hosted by a German provider. Our team can now work productively with AI without resorting to external tools or shadow AI. Working together is straightforward, and questions are answered quickly and clearly. With WZ-IT we also have a partner at our side for further AI projects. (translated)
Carl-Frederic KornHead of E-Commerce & Online Marketing, Krüger Dirndl GmbH
Knowledge Management & AI
The knowledge from many years in the solar industry lived in the head of our former managing director, and it still lives in the heads of our employees today. We wanted to secure it, not as a folder full of notes, but in a way that lets us ask questions in daily work and get reliable answers. WZ-IT advised us from the start and weighed different approaches and knowledge management systems against each other with us before we decided. What was built is a custom knowledge base with an AI assistant that brings together conversations, notes, documentation and manuals: from the platform through onboarding our team to support in operation. Conversations are transcribed and facts are extracted; we review and correct everything ourselves before it is stored, because we do not want to trust an AI blindly. The system runs in our own environment, and the data stays with us. What surprised us was how efficient it was: WZ-IT used existing open-source software and extended it specifically for our requirements, instead of selling an expensive custom build from scratch, as is often usual with IT service providers. Timo and Robin understood our actual problem, preserving knowledge during the generational handover, and implemented it step by step: from the proof of concept through setup and validation with our own data to productive operation, explaining everything clearly along the way. Next, we want to extend this to the knowledge of our employees, so that this does not depend on individual people either.
Joshua von AmelnManaging Director, solarwerkstatt berlin GmbH
International
WZ-IT designs, develops and operates infrastructure and software for clients in Germany and internationally. We deliver projects remotely and continue supporting them in ongoing operations after go-live.
Non-binding enquiry
Tell us the model, quantisation, context, concurrent usage and required application layer. We assess whether 24 GB, 96 GB or a custom architecture fits.
The conversation is free and non-binding.
WZ-IT's advice on our Azure migration was technically sound and completely non-binding right from the intro call - we took away a great deal.
Jakob ÖschlbergerInno7 GmbH
WZ-IT's advice on our Azure migration was technically sound and completely non-binding right from the intro call - we took away a great deal.
