One box. Full AI capability. Nothing leaves the building.
A turnkey, physically self-contained AI appliance, hardware, software, and support in one annual subscription. Runs air-gapped or on a closed local network. Sold in three simple tiers by team size.
Three steps. No jargon.
Install
The Box arrives ready to use. Connect it to your local network, wired or Wi-Fi, or run it fully disconnected, with no network at all.
Run
Your team uses AI for drafting, research, summarizing, and analysis. Everything happens on the device itself.
Stay current
Updates and support are included in your annual subscription. Nothing for your team to manage.
Physical control, not just a promise in a privacy policy.
No mandatory internet
Run fully disconnected, or LAN-only so your whole office can use it with zero external exposure.
Never trained on your work
Nothing is ever used to train an outside model, because your work never leaves the device.
In your building
The hardware sits in your office, under your access rules, not a vendor's data center.
Five capabilities, all on-device. All in one appliance.
Everyday writing, reasoning & task automation
Drafting, summarizing, answering questions, plus carrying out small multi-step workflows on its own. Not just answering a prompt: checking a document against a rule, pulling numbers from a report, drafting the follow-up.
Powered by · open models including gpt-oss-20b, Nemotron 3 Nano, and Gemma 4.
Document intelligence at real scale
Search and answer across huge internal collections, contracts, case files, lab notebooks, research papers, technical manuals. Hundreds of thousands of pages, with each answer pointing back to the source it came from.
Powered by · Ingestion powered by NVIDIA Nemotron Parse; retrieval via modern open embeddings.
Vision & image recognition
Read and describe scanned documents, lab or diagnostic images, engineering diagrams, and photos. Pulls structured information out of messy, real-world PDFs and forms, not just clean digital text.
Powered by · open vision-language models such as Qwen2.5-VL and InternVL3.
Voice: transcription & speaker recognition
Turn spoken conversation into accurate written text and tell speakers apart. Useful for meeting notes, visit documentation, and call review, entirely on-device.
Powered by · Whisper-class transcription with modern diarization models.
Long-context understanding
Some models the Box can run read and reason across extremely long material in a single pass, in some cases the equivalent of a very long book. A full case file, research set, or technical manual, considered at once.
Powered by · Long-context reasoning via DeepSeek R1-Distill and Gemma 4 variants.
Curious what's actually running under the hood? Read our plain-English guide to the models →
Pick by headcount, not by spec sheet.
Every tier includes hardware, software, setup, and support in one annual subscription , no separate bills, no per-use charges, no surprise invoice.
Box: Studio
Up to 5 people
- Ideal for
- Solo practitioners, small labs, boutique firms
- Deployment
- Air-gapped or closed local network, wired or Wi-Fi
- Support
- Standard annual support & updates
- Onboarding
- Guided self-setup + remote onboarding session
- Consulting
- Starter session included
Box: Team
Up to 15 people
- Ideal for
- Growing teams, small clinics, scale-up R&D groups
- Deployment
- Air-gapped or closed local network, wired or Wi-Fi
- Support
- Priority support + quarterly check-ins
- Onboarding
- On-site or remote onboarding session
- Consulting
- Ongoing quarterly session
Box: Custom
Up to 100 people
- Ideal for
- Multi-department orgs, distributed sites
- Deployment
- Air-gapped, closed network, or multi-site
- Support
- Dedicated account manager + custom SLA
- Onboarding
- Full on-site deployment & staff training
- Consulting
- Custom consulting package
Tell us about your team.
Pricing depends on your team size and setup. Share a few details and we'll get back to you with a quote, no calendar bookings, no sales runaround.