First units shipping Q2 2026

Own Your AI.
Keep Your Data Private.

Turnkey on-prem AI inference boxes for regulated SMBs. Run sensitive workloads locally with full audit trails and compliance guardrails — while intelligently routing everything else to AWS Bedrock for speed and cost savings.

On-prem deploymentHybrid Bedrock routingImmutable audit logsStarts at $39k

Hardware we're shipping now

Purpose-built AI inference boxes

Air-cooled, office-friendly designs. No liquid cooling, no data center required. Built with Supermicro and Dell hardware, powered by NVIDIA and AMD GPUs.

Entry Box

Perfect for First Pilots & Small Teams

$39k–$45k

Pilot price — includes setup + 1-day vibe-coder workshop

GPU4× RTX PRO 6000 Blackwell or 4× AMD MI300X
VRAM384–768 GB
Power1.2–1.8 kW (standard office outlet)
Performance25–60 tok/s single-user · 800–2,000+ aggregate

Form factors

Under-desk tower2U short-depth closet server6U wheeled flight-case

Small law firms, dental practices, K-8 schools, or solo practitioners

Best for teams of 10–30 users

Recommended

Mid Box

Production Workhorse for Growing SMBs

$89k

Pilot price — includes setup + 2-day workshop + first-app sprint

GPU4× NVIDIA H200
VRAM564 GB
Power1.6–2.2 kW
Performance40–80+ tok/s single-user · 1,500–4,000+ aggregate

Form factors

4U rackmount8U reinforced flight-case

Mid-size law firms, healthcare clinics, school districts, or manufacturers

Best for teams of 20–60 concurrent users

Why teams choose GenTrellis

More than cloud AI. More than raw hardware.

GenTrellis combines on-prem sovereignty with intelligent cloud routing, hands-on enablement, and compliance guardrails that regulated SMBs actually need.

True Privacy

Sensitive workflows — contracts, patient notes, student data, PLC code — never leave your network. Your data stays on infrastructure you control.

Smart Economics

Hardware is paid once. The intelligent router fills spare GPU capacity automatically. Most customers break even in 4–9 months vs. pure cloud.

See the full comparison

Built for Your Team

We don't just ship hardware — we teach your vibe coders how to build secure internal apps using smol-agent, RAG, and guardrails.

Developer enablement

Future-Proof

Start with an Entry box and add Mid or Enterprise units as you grow. Seamless Ray clustering. Bedrock resale gives flexibility.

“GenTrellis sees nothing — ever.”

Your data. Your policies. Your audit trail.

Protection levels

Local-first is the foundation.
Protection levels are the differentiator.

Some workflows need more than privacy. They need rules about what the AI can see, say, retrieve, and do. Choose the right level for each workflow.

Source-grounded with PII detection and content policy

What is our refund policy for students who withdraw mid-semester?

Based on the student handbook, students who withdraw before the census date receive a full tuition refund. After census date, refunds are prorated: 75% in weeks 3-4, 50% in weeks 5-8, and 0% after week 8. The withdrawal form must be submitted to the Registrar's office.

ProtectedVerified against 12 documents · PII check passed · Content policy clear

How it works

From deposit to production in weeks

1

Reserve your pilot box

Choose Entry or Mid hardware. 50% deposit secures your Q2 2026 delivery slot.

2

We set up and train your team

On-site or remote setup, data connection, and a hands-on vibe-coder workshop to build your first apps.

3

Run AI on your terms

Sensitive work stays local with protection levels. The router handles the rest through Bedrock automatically.

What's included

Every pilot comes ready to run

Not just hardware. Setup, training, software stack, and support — everything your team needs to ship production AI apps.

On-site or remote setup

We configure your box, connect your data, and validate everything works before we leave.

Vibe-coder workshop

1–2 day hands-on workshop where your team builds production apps on your own data using smol-agent and local RAG.

Learn more about workshops

Intelligent Router

Sensitivity-based tiering routes sensitive work locally and everything else to Bedrock — automatically.

Pre-loaded AI stack

vLLM inference, smol-agent multi-agent harness, local RAG, NeMo Guardrails, and policy engine — ready to go.

Developer details

30 days priority support

Priority support plus audit-ready logs from day one. We're available while your team ramps up.

50% deposit secures delivery

Reserve your Q2 2026 delivery slot with a 50% deposit. Balance due on delivery.

For developers & vibe coders

Not just hardware — we teach your team to build

GenTrellis exposes an OpenAI-compatible API your team can vibe-code against. Unlike Claude Code, your solutions run against local endpoints with protection levels, RAG, and multi-agent crews built in. We include hands-on workshops to get your team shipping production apps on day one.

FAQ

Common questions

Reserve your pilot

Pilot slots for Q2 2026 are limited

Reserve your pilot box starting at $39k with a 50% deposit. We'll set up your hardware, connect your data, and train your team.

No spam. We'll only reach out to schedule your pilot setup and discovery call.

Air-cooled, office-friendlySOC 2 readiness underwaySample BAAs & FERPA/HIPAA guidanceImmutable local audit logs