# We don't just sell boxes. We get your AI running.

> OnPrem World · https://onpremworld.com/en/solutions · Last updated: 2026-10-16

Hardware, models, a chat interface, accounts and support, planned together. Try it on your own documents before you commit.

## Starter

Small teams trying private AI. One DGX Spark or RTX workstation in your office with Open WebUI, the right models and user accounts set up.

- Hardware, installation and setup
- Model selection and tuning for Chinese and English
- One training session for your team

## Team

A department sharing one system. A multi-GPU RTX PRO 6000 workstation for document work, coding assistants and your knowledge base.

- Accounts, permissions and usage logs
- Knowledge-base (RAG) integration with your files
- Monitoring and remote support

## Enterprise

Company-wide private AI. DGX Station or GPU servers for many users and models at once, and fine-tuning on your data.

- Architecture plan and a proof of concept on your data
- Security and internal-control documents
- Support contract with agreed response times

## How a project runs

1. **Talk**: What you want to run, for how many people, and how sensitive the data is.
2. **Assess and trial**: A proof of concept on your own documents, usually two to four weeks.
3. **Quote and order**: Configuration, schedule, warranty and how long the price holds, in writing.
4. **Set up**: Install, configure, connect to your systems, sign off.
5. **Train and support**: We teach your team, then we look after the system.

## Frequently asked questions

### Why run AI on your own hardware instead of a cloud service?

Your documents and prompts never leave the building, which matters for personal data, finance, health and legal work. Costs are fixed instead of per-token, and you choose which models run.

### Which models do you set up?

Open-weight models chosen for your use and your hardware, for example gpt-oss, Qwen, Gemma or Llama, tested on your own documents in Chinese and English. Some public-sector buyers cannot use certain models; we check this with you first.

### How much does a project cost?

It depends mostly on the hardware. Taiwan shelf prices for the hardware we use are on our prices page; setup and support are quoted per project.

### Do you offer a trial?

Yes. For companies we usually start with a short proof of concept on your own data before you buy.

Request a quote: https://onpremworld.com/en/quote · alexma@painpoint-ai.com