Hardware · The Sovereign AI Box

Your AI as capital, not rent.

One appliance on your LAN serves unlimited-token AI to everyone: vLLM-class serving, MoE models, per-app LoRA.

Own it — from $50← Back to the suite

SB-30 · ~$4,400

30 GB · 5–15 users. 24–32B dense at Q4, or a ~30B MoE.

SB-50 · ~$6,900

50 GB · 15–30 users. 70B Q4 fits tight; 64K+ context.

SB-75 · ~$11,000

75 GB · 30–60 users. 70B Q4 comfortably, long context.

SB-150U/X · flagship

150 GB · 60–200 users. 100–140B MoE, frontier-adjacent.

Sovereign AI Box — SB-50 · your LAN
Serving · 22 seats
unlimited tokens

Five-box lineup (SB-30→150X)

Unlimited tokens, throughput-sized

vLLM-class batched serving

Shared base + per-app LoRA

Grammar-constrained tool calls

mDNS provisioning

Sovereign Tuning on your data

BYO hardware + trade-in

1

Provision

mDNS LAN discovery brings the box online; enroll against your key.

2

Serve

The box serves the local model to every terminal — unlimited tokens.

3

Tune

Sovereign Tuning trains LoRA on the box; your data never leaves.

Trade your existing box in and get a 15% credit on your new one.