On-prem · Open-source AI

Your own private AI, on your own hardware.

SunLabs installs open-source large language models on a server inside your building — a private ChatGPT-style assistant that reads your documents and answers your team, with no cloud, no per-token fees, and no data ever leaving your walls.

See the packages Request a quote
What you get

Everything in a single, private install.

Every package is a complete, self-contained system. We supply and configure the hardware, the model, and the software — then train your team to use it.

The hardware

A GPU server sized to your needs, racked and configured on-site. You own it — no monthly cloud bill.

Open-source models

Leading open models — Llama, Mistral, Qwen, DeepSeek — running fully offline, so nothing is sent to a third party.

Private chat + API

A clean chat interface for your staff plus an API your other tools can call — all on your local network.

Document pipeline (RAG)

Point it at your files and it answers from your own documents, contracts, and records — with citations.

Air-gapped privacy

Can run with no internet connection at all. Privileged and regulated data stays inside the building, full stop.

Setup + team training

We install, tune, and hand it over with hands-on training so your team is productive from day one.

Packages

Three ways to start.

Packages scale by model size and how many people use it at once — from a single workstation to a full enterprise rack. Every tier is a one-time build; exact pricing depends on the hardware and model you need.

Starter Node
Single-GPU workstationFor a small team or one department
A compact workstation that runs capable mid-size models — great for a first private AI deployment.
  • 1 GPU (24–48GB VRAM)
  • Runs 8B–34B open models
  • Private chat UI + local API
  • Document pipeline (RAG)
  • Setup, tuning & team training
Request a quote
Most Popular
Team Server
Multi-GPU serverFor a whole team, in daily use
A server-grade build that runs large 70B-class models with room for many users and a bigger document library.
  • 2 server-grade GPUs
  • Runs up to 70B open models
  • RAG over your full document set
  • Role-based access for the team
  • Setup, tuning & training included
Request a quote
Enterprise Rack
Multi-GPU H100-class nodeFor heavy, org-wide workloads
Maximum capability and reliability — the largest open models with redundancy and a governed data pipeline.
  • Multi-GPU H100-class node
  • Runs the largest open models
  • High availability & redundancy
  • Full data ingestion & governance
  • Dedicated onboarding & support
Request a quote

Not sure which fits? Tell us your data, your team size, and your privacy requirements — we'll spec the right build and quote it.

Why on-prem

Own your AI instead of renting it.

Cloud AI means your data leaves your building and you pay per use, forever. An on-prem install is a one-time investment: your models, your hardware, your data — private by design, with predictable costs and no vendor lock-in.

Get a package spec'd for your business.

Tell us what you're working with and we'll recommend the right hardware and model, then send a detailed quote.