Ikioma Mini
The desk-sized entry point — two linked nodes, one private stack.
from €50,000
View →
Made to order · 2–8 weeks
Configure yours →
Ikioma Mini
The desk-sized entry point — two linked nodes, one private stack.
From €50,000
Hardware, delivered and deployed
Support & optimization
€8,000/year
Year 1 total
€58,000
[ why own it ]
01 · physics
Private by physics.
Your data never leaves the floor. Air-gappable, no egress, no telemetry — the trust boundary has a serial number you can point at in an audit.
02 · metering
Pay once, infer forever.
Own the machine and agents run around the clock. No per-token meter, no overage invoices — you pay for electricity, not tokens.
03 · ownership
Yours to keep.
Hardware is a capital asset, not a subscription line. Weights, tunes, and stack stay yours — perpetually, with trade-in credit on next-gen silicon.
01 · the desk
The whole system, two small cubes.
Two 150 mm nodes linked into one machine — 256 GB of unified memory and a tuned 32B model, quiet enough to sit on your desk next to your monitor. No rack, no fan noise, no data center.
2× 150 mm cubes
two linked nodes
256 GB unified
unified memory, one pool
35dB
quiet enough for a desk
02 · your team
A small team's worth of agents.
Eight concurrent agents, 128k tokens of context, and models tuned for the work your team actually does — enough inference to start building without standing up infrastructure.
8
concurrent agents
128ktok
context
32 B · 4-bit
tuned model, 4-bit
03 · the stack
The full stack, minus the rack.
The same model and inference stack as every Ikioma — Conduit, a drop-in OpenAI/Anthropic API, signed audit log, sandboxed tools. Start here; trade in toward Ikioma when the workload outgrows the desk.
same
stack as every Ikioma
/v1
drop-in API
trade-in
credit toward Ikioma
04 · economics
Own your inference from day one.
Private inference is a capital asset, not a subscription line — no per-token meter, no overage invoices. Most customers break even versus equivalent cloud spend within 4–7 months; security updates run 10 years, in writing.
4–7mo
typical break-even vs cloud
10yr
security updates, in writing
0meter
no per-token billing
[ heritage ]
Built in the lineage of instruments, not platforms.
Assembled in Yokohama and Eindhoven. Boards burn in for 96 hours before they leave the floor. Every system carries a serial, a calibration sheet, and a name etched on the chassis — because the people who built it stand behind it for a decade, not until next quarter's earnings call.
96h
burn-in per system
before it leaves the factory floor
before it leaves the factory floor
2
assembly sites
Yokohama · Eindhoven
Yokohama · Eindhoven
10yr
security update commitment
in writing
in writing
[ the system ]
Hardware spec
Silicon
2× IKM-1, linked
Memory
256 GB unified — two nodes, one pool
Params
32 B · 4-bit
Network
2-node link · 200 Gb/s
Power
2× 240 W · silent
Dimensions
2× 150 mm cubes
Weight
~5 lbs
lead_time
Made to order · 2–8 weeks
Model & inference stack Included
Preinstalled and pre-validated: the ikioma inference runtime, your choice of optimized open models, and the production middle layer (audit log, RBAC, evals) configured out of the box.
- ✓OpenAI / Anthropic API compatibility
- ✓Sandboxed tool execution
- ✓Signed audit log
- ✓Runs on-premises — your hardware, your premises
| Model | Params | Context | tok/s | Latency | Notes |
|---|---|---|---|---|---|
| ikioma-32B | 32B | 128k | — | — | TBD — burn-in benchmark |
Loading 3D view…
Drag to rotate · Scroll to zoom
Cluster
GB10 telemetry · simulated
GPU temp—
GPU clock—
Unified mem—
Blowers—
System power—
the hardware
The DGX Spark, exploded.
Every part of the SFF node you can pull apart — X-ray the shell to see the GB10 and 8× LPDDR5X, drag the explode slider for the teardown order, or switch on airflow and telemetry. Drag to orbit, scroll to zoom.
Illustrative 3D approximation — not manufacturing CAD. Schematic dimensions only.
[ which ikioma ]
Three versions, one stack.
Mini for the desk, Ikioma for the team, Pro for the fleet. Every version ships the same model and inference stack — the scale is what changes.