Skip to main content
← Sovereignty
Posture / Owned

Inside your perimeter. On a box you can touch.

Owned puts the reasoning substrate on a 1331 appliance — Mini, Studio, or Rack — installed inside your network boundary. Open-weight models, signed offline update bundles, and no third-party model provider in the inference path.

/ What ships with Owned

Inside the boundary

Racked inside your assessed network segment. No required outbound connectivity for inference, retrieval, or audit.

Signed offline updates

Model weights and runtime updates ship as signed bundles. The appliance never has to call home.

Open-weight models, US supply chain

Llama, Qwen, Mistral, DeepSeek, gpt-oss — pinned versions you choose, not a moving SaaS endpoint.

Hardware-rooted keys

Secure-enclave-rooted keys. Per-workload scoping; rotation is a first-class event in the audit log.

/ Appliance Line

Real machines, racked and configured by a real team.

1331 Mini

$1,499
Apple Silicon · 24GB unified memory · 512GB SSD

Solo founders, small teams, internal agents up to ~13B class models.

  • Up to 13B parameter LLMs at 30+ tok/s
  • ~70W typical draw
  • Fanless under load
  • Wired Ethernet + Wi-Fi 6E
Most popular

1331 Studio

$3,299
Apple Silicon Pro · 64GB · 2TB SSD

Production agents, RAG over company data, multi-user dev teams.

  • Up to 70B class models (quantized)
  • Concurrent multi-tenant inference
  • Local vector store + embedding pipeline
  • Ships with agent runtime preinstalled

1331 Rack

From $9,500
1U workstation · NVIDIA RTX 5090 or dual-GPU options

Heavy throughput, fine-tuned models, regulated industries on-prem.

  • Llama 4, DeepSeek V3.2, gpt-oss large
  • Hot-swappable storage
  • Optional redundant PSU
  • Rack-and-stack install in NJ / NYC metro
/ Why a real box, not a hyperscaler region

Apple Silicon and 1U workstations beat 'EU-region SaaS' on the only test that counts.

An FSO, a CMMC assessor, a NERC CIP auditor, and a Schrems II-aware DPO all want the same thing: a boundary they can draw on a diagram and a vendor relationship that does not silently extend that boundary the moment a prompt leaves the building. Owned answers that without a contract argument.

Unified memory

The whole model lives in RAM the GPU can read directly — no PCIe round-trips, no quantization gymnastics for 70B-class workloads.

Power-efficient

Runs from a standard wall outlet. Mini and Studio go under a desk or in a closet; Rack drops into a standard 1U slot.

Imaged, racked, supported

Every appliance is benchmarked and burn-in tested before it ships. Spares ship with each engagement.

/ When to pick Owned

Strong fit

  • Defense, DIB primes, ITAR-touching subs
  • Law firms with GC-enforced AI policy
  • Hospitals running PHI through ambient documentation
  • Utilities / NERC CIP-regulated operators
  • Any buyer whose regulator will ask to see the box

Pick a different posture

  • Workloads that legitimately need elastic burst across regions
  • Teams without on-site IT or a remote-hands relationship
Request a BriefingCompare posturesAudit & Controls