Superchat Box Lite
Private chat for a small team or one website.
$5,500 one time, zero token cost
- Genie Lite
- Genie Chat
- Live Support
Superchat Box
Your company's AI. In two sizes.
Box Lite for private chat. Box Pro for chat, voice and image. Both run offline, both stay in your building, neither sends you a token bill.
Get one for your website:
Hi! Ask me anything about the Superchat Box.
Superchat Box Pro
Chat, voice and image for the whole office.
$20,000 one time, zero token cost
- Genie Lite + Genie Pro
- Chat and Live Support
- Voice (Orb)
- Image
Both ship with Ubuntu and upgrade to Genie OS from March 2027 · Quoted per deployment · Compare models
Superchat Box Pro
Get the highlights.
Chat, voice and image. Genie Lite and Genie Pro, with the Radeon 8060S graphics running voice and image.
Room for a big model. Memory shared by the CPU, graphics and NPU.
128 GBunified memory, up to 96 GB for AI
Ports where you need them. USB4 and 10 Gbps USB on the front. 10 Gb Ethernet on the back.
Made to sit in plain sight. Brushed aluminium, a honeycomb grille, 180 mm square.
Faster than you can read. Genie Pro, measured on the box.
48tokens per second1
Ubuntu today. Genie OS next. Every Superchat Box upgrades to Genie OS from March 2027.
2027Genie OS arrives in March
Inside Box Pro.
180 by 180 by 78 millimetres. About the size of a hardback book. Slide off the side and there is no graphics card. One processor does it all. 128 GB of memory sits right beside it. That is what lets Genie Pro run here.
- AMD Ryzen™ AI Max+ 39516 cores, Radeon 8060S, 50 TOPS NPU
- 128 GB LPDDR5XShared memory, up to 96 GB for AI
- Vapour chamber, dual fansUp to 140 W, kept cool
- 4 × M.2 NVMe1 TB installed, up to 8 TB each
- Back panel10 GbE, 2.5 GbE, USB4, 4 × USB, HDMI 2.1, DisplayPort 2.1
- Front panelUSB4, 2 × USB 10 Gbps, SD card
Illustration. Board layout simplified.
Capacity
Enough for the whole office.
People ask, read, think and ask again, so a team never asks all at once. One box answers four people at full speed at the same moment, and keeps an office of 50 to 100 people moving through an ordinary working day.4
4answers at once
Each at about 22 tokens per second, four times faster than you read
75tokens per second
Total output, shared by everyone on the box
50–100people
A typical office for everyday chat. Bigger teams add a second box.
Superchat Box Lite
Private chat. Small enough to hide.
A 13 cm box that runs Genie Lite for your team and your website. It mounts behind a monitor or on a wall, and like Box Pro it works with the internet switched off.
Model
Genie Lite
Fast, private answers for everyday questions and website conversations.
Apps
Chat and Live Support
Genie Chat for your team and the assistant on your website, installed and ready.
Hardware
Ryzen AI 9 HX 470
12 cores, 32 GB memory, 1 TB storage, two 2.5 Gb network ports. Wall-mount bracket included.
Software
Two Genie models. Pick your box.
Box Lite runs Genie Lite. Box Pro runs both, and adds voice and image. Everything is installed before the box reaches you.
Fast · Box Lite and Box Pro
Genie Lite
Everyday questions, website conversations and spoken replies, in an instant.
120B-equivalent · Box Pro
Genie Pro
Reasoning, long documents, analysis and code.
Operating system
Ubuntu today. Genie OS next.
Every Superchat Box ships with Ubuntu and Genie ready to use. From March 2027, every box will be upgraded to Genie OS, our own operating system built around private AI.
Available now
Ubuntu
The trusted Linux your IT team already knows, with Genie pre-installed.
From March 2027
Genie OS
Every Superchat Box will be upgraded to it.
Cost
Pay once. Stop counting tokens.
Cloud AI charges for every word it reads and writes, so the bill grows as your team uses it more. The box has a price, not a meter.
How they use AI
Cloud API bill, per year$0
Per-token bill on the box$0
Estimate for 0 tokens a year at a blended $10 per million tokens and 22 working days a month. A Superchat Box is a one-time $5,500 (Lite) or $20,000 (Pro).
Compare
Which box is right for you?
| Feature | Superchat Box Lite | Superchat Box Pro |
|---|---|---|
| Best for | Chat for a small team or one website | Chat, voice and image for the whole office |
| Genie models | Genie Lite | Genie Lite and Genie Pro (120B-equivalent) |
| Genie Chat | Yes | Yes |
| Genie Live Support | Yes | Yes |
| Voice (Genie Orb) | No | Yes |
| Image: create and read images | No | Yes |
| Processor | AMD Ryzen AI 9 HX 470, 12 cores | AMD Ryzen AI Max+ 395, 16 cores |
| Graphics | AMD Radeon 890M | AMD Radeon 8060S, 40 graphics cores (runs voice and image) |
| Memory | 32 GB | 128 GB unified, up to 96 GB for AI |
| Storage | 1 TB, 2 × M.2 slots | 1 TB, 4 × M.2 slots |
| Network | 2 × 2.5 GbE | 10 GbE + 2.5 GbE, Wi-Fi 7 |
| Size | 132 × 132 × 50.5 mm | 180 × 180 × 78 mm |
| Price, one time | $5,500, no token bill after | $20,000, no token bill after |
| Measured speed | Not yet published | 48 tokens/s writing, 1,175 reading |
Tech specs.
Superchat Box Pro
Processor
- Processor
- AMD Ryzen™ AI Max+ 395
- Architecture
- Zen 5, 16 cores / 32 threads
- Clock speed
- 3.0 GHz base, up to 5.1 GHz boost
- Graphics
- AMD Radeon™ 8060S, 40 compute units; runs voice and image
- NPU
- AMD XDNA™ 2, up to 50 TOPS
- AI compute
- Up to 126 total TOPS; up to 60 FP16 TFLOPS on the GPU
- Processor power
- Up to 140 W in this chassis
Memory and storage
- Memory
- 128 GB LPDDR5X-8533, soldered
- Graphics memory
- Up to 96 GB of the 128 GB can be assigned to the GPU
- Storage
- 1 TB NVMe SSD installed
- Expansion
- 4 × M.2 2280 NVMe slots (2 × PCIe 4.0 ×4, 2 × PCIe 4.0 ×2), up to 8 TB each
Connectivity
- Front
- Power button, 2 × USB 3.2 Gen 2 (10 Gbps), 1 × USB4, SD card slot
- Back
- 10 Gigabit Ethernet, 2.5 Gigabit Ethernet, 1 × USB4, 4 × USB 3.2 Gen 1 (5 Gbps), HDMI 2.1, DisplayPort 2.1, 3.5 mm audio, DC in
- Wireless
- Wi-Fi 7, Bluetooth 5.4
- Displays
- Up to four, at up to 8K 60 Hz
Design
- Size
- 180 × 180 × 78 mm
- Weight
- About 2.3 kg
- Cooling
- Vapour chamber, copper heatsink and dual fans
- Power supply
- External adapter, 19.5 V 11.8 A
Capacity
- Answers at once
- 4 at full speed, about 22 tokens per second each
- Total output
- About 75 tokens per second, shared by everyone using the box
- Typical team
- About 25 people chatting at the same time: an office of 50 to 100 for everyday use
Operating system
- Available now
- Ubuntu, with Genie pre-installed
- From March 2027
- Genie OS. Every Superchat Box will be upgraded to it.
Software
- AI models
- Genie Lite and Genie Pro, pre-installed
- Genie Lite
- Fast model for chat, website support and voice
- Genie Pro
- 120B-equivalent model for reasoning, documents and code
- Included apps
- Genie Chat, Genie Live Support, Genie Orb (voice), image creation and reading
- Operation
- Runs fully offline. No account, licence server or cloud link required to answer.
- Updates
- Delivered by Superchat on a USB drive or over the network, your choice
Warranty
- Hardware warranty
- 1 year
Superchat Box Lite
Processor
- Processor
- AMD Ryzen™ AI 9 HX 470
- Architecture
- 12 cores / 24 threads
- Graphics
- AMD Radeon™ 890M
Memory and storage
- Memory
- 32 GB LPDDR5, soldered
- Storage
- 1 TB NVMe SSD installed
- Expansion
- 2 × M.2 2280 NVMe slots; MCIO expansion connector
Connectivity
- Front
- Power button, reset, 3.5 mm audio, 2 × USB 3.0, MCIO
- Back
- 2 × 2.5 Gigabit Ethernet, USB4, HDMI 2.1, DisplayPort 2.1, USB 3.0, USB 2.0, DC in
Design
- Size
- 132 × 132 × 50.5 mm
- In the box
- Power adapter and cable, wall-mount bracket and screws
Operating system
- Available now
- Ubuntu, with Genie pre-installed
- From March 2027
- Genie OS. Every Superchat Box will be upgraded to it.
Software
- AI model
- Genie Lite, pre-installed
- Included apps
- Genie Chat, Genie Live Support
- Operation
- Runs fully offline. No account, licence server or cloud link required to answer.
Warranty
- Hardware warranty
- 1 year
Platform figures are the manufacturer's. Processor figures are AMD's published specifications; TOPS and TFLOPS are peak theoretical values. Box Lite speed and capacity have not been measured yet.
Support and documents.
Box Pro user manual
Setting up, first start, ports, care, troubleshooting and safety. PDF.
DownloadBox Lite user manual
Setting up, mounting, ports, care, troubleshooting and safety. PDF.
DownloadTalk to an engineer
Sizing, installation in your office or rack, and connecting your own systems.
Contact usFor security teams
How the box stays offline, what it stores and how updates reach it.
Read more- Genie Pro on a production Superchat Box, measured 8 November 2026: median of five runs of 256 output tokens, one request at a time, temperature 0. Writing 47.6–48.1 tokens per second; reading measured on a 2,540-token document. Raw results.
- Total time to read a 2,540-token contract and return a two-sentence answer, median of three runs.
- Genie Pro measured on a Superchat Box, 8 November 2026, with speculative decoding on (output is unchanged by it). Reading speed: 238 words a minute, the average silent reading rate for English non-fiction (Brysbaert, 2019, Journal of Memory and Language), converted at about 0.75 words per token.
- Capacity measured 8 November 2026 on a Superchat Box running Genie Pro: 1 to 16 different requests of about 180 words sent at the same moment. Four answers stream at full speed together (about 22 tokens per second each); total output levels off at about 75 tokens per second, and further simultaneous requests wait their turn. Office sizing assumes about a quarter of staff chatting at the busiest moment, each asking a question every few minutes; long documents and coding agents need more. Raw results.
- Processor figures are AMD's published specifications; TOPS and TFLOPS are peak theoretical values. Platform figures are the manufacturer's. AMD, Ryzen and Radeon are trademarks of Advanced Micro Devices, Inc.