Own 1 · Personal AI Computer
Fast at your desk.
Global when it counts.
A silent, AI-optimized computer that runs open models locally, keeps your data on device by default, and extends to OwnCloud when a workload needs the world’s compute.
$2,999
Starting price
Near-zero
AI running cost
OwnCloud
Cloud on tap
Overview
A personal computer built around local intelligence.
Own 1 is designed to run open language, vision and voice models on your desk — with on-device privacy, AI-native security, native agents, a machine wallet and a direct path to OwnCloud when a workload needs more.
01
Local-first
Models, agents and data stay on device by default.
02
AI-optimized
Silent thermals tuned for continuous inference.
03
Hybrid
OwnCloud on tap when a workload needs global compute.
04
Ownable
Rollback recovery, open software, no forced services.
Your model. Your data. Your machine.
Own 1 · Personal AI Computer
Fast at your desk. Global when it counts.
A silent, AI-optimized computer that runs open models locally, keeps your data on device by default, and extends to OwnCloud when a workload needs the world’s compute.
$2,999
Starting price
Near-zero
AI running cost
OwnCloud
Cloud on tap
Market position
See where Own 1 sits in the market — before you learn about the product.
A category snapshot against Mac mini, premium mini PCs and DGX-class workstations. Every advantage below maps to a benchmark or spec disclosed lower on this page.
| Category | Own 1 | Mac mini | Premium Mini PC | DGX-Class |
|---|---|---|---|---|
| Starting Price | $2,999 | Lower | $1k–$2.5k | $3,999+ |
| AI Running Cost | Near-zero locally | Near-zero (manual) | Near-zero (manual) | Near-zero |
| Cost Saving | High | Medium | Medium | High |
| AI Speed | AI Optimized | Strong | Strong | Highest |
| Privacy | Local-first | Depends | Depends | Local |
| Security | AI-native | OS Security | OS Dependent | Enterprise |
| Open Source Models | ✅ | Manual | Manual | ✅ |
| Stablecoin Payments | ✅ | — | — | — |
| Smart Contracts | ✅ | — | — | — |
| AI Wallet | ✅ | — | — | — |
| AI Marketplace | ✅ | General Apps | No | Limited |
| Local Model Hosting | ✅ | Manual | Manual | ✅ |
| AI Agents | ✅ | Manual | Manual | Enterprise |
| OwnCloud Hybrid | ✅ | — | — | Vendor |
| Offline AI | ✅ | Possible | Possible | Yes |
| Rollback Recovery | ✅ | Standard | OS | Enterprise |
Estimated · Reference specs and pricing bands, 2026 snapshot
Device specs
Choose your Own 1 — Pro or Max.
Two configurations, one architecture. Pro is tuned for developers and daily AI power users. Max doubles storage and unlocks boost headroom for heavier workloads and long-running agents.
| Specification | Own 1 Pro | Own 1 MaxRecommended |
|---|---|---|
| Price | $2,999 USD | $3,399 USD |
| Positioning | For AI power users, developers and builders | Maximum capacity for heavier workloads and long-running agents |
| CPU | Intel Core Ultra 9 · 24 cores | Intel Core Ultra 9 · 24 cores |
| Unified RAM | 128 GB | 128 GB |
| NVMe SSD | 4 TB | 8 TB |
| On-device AI | 99 TOPS | 99 TOPS + boost |
| Operating system | OwnXOS · secure enclave | OwnXOS · secure enclave |
| Warranty | 3-year warranty · lifetime updates | 3-year warranty · lifetime updates |
Reference · ownx.co/Own1 preorder configuration
Benchmarks
Every benchmark visible. No tabs. Full-width charts.
Nine benchmark groups covering cost saving, text, coding, image, video, speech, voice, music and an overall AI radar. All values are DEMO DATA · NOT MEASURED until the Own 1 launch evidence pack (hardware, model file, hash, quantization, runtime, ten repeated runs, raw JSON) is published.
Benchmark · 01
Cost saving.
Local vs Cloud (monthly)
Modelled monthly cost for a moderate AI workload run locally vs on hosted APIs.
Loading data…
Assumes 8h/day AI use · reference workload mix
API savings by workload
Percent saved by running each modality locally on Own 1 vs equivalent hosted API.
Loading data…
Modelled at reference API prices
Break-even vs cloud
Months to pay back Own 1 versus equivalent hosted AI subscriptions.
Loading data…
Straight-line payback · excludes electricity
Hybrid rule
Local when possible. OwnCloud when it’s cheaper or faster.
Own 1 routes each workload to the cheapest suitable target automatically — on-device first, OwnCloud second, hosted API only if you approve it.
Benchmark · 02
Text & Reasoning.
Tokens per second
Sustained generation throughput on a 13B-class open chat model.
Loading data…
Ref hardware · 8-bit quant · median of 10 runs
Time to first token
Latency from prompt submit to first streamed token.
Loading data…
Same prompt · warm cache · median of 10 runs
Long-context performance
Maximum usable context window at interactive speed.
Loading data…
Interactive threshold ≥ 20 tok/s
Largest supported model
Largest open model that fits in memory at 4-bit.
Loading data…
4-bit quant · single-GPU/APU · no offload
Benchmark · 03
Coding.
Repository indexing
Time to index a 50k-file monorepo for local retrieval.
Loading data…
Local RAG · cold cache
Code generation
Sustained tokens per second on a 7B code model.
Loading data…
Function-completion prompts · median of 10
Bug fixing
Pass rate on a curated bug-fix regression set.
Loading data…
Internal bug-fix set · pass@1
HumanEval
pass@1 on the standard HumanEval coding benchmark.
Loading data…
Public HumanEval · pass@1
Benchmark · 04
Image generation.
Seconds per image
1024×1024 generation, 30 steps, SDXL-class open model.
Loading data…
Same model file · median of 10
Images per hour
Continuous generation throughput at fixed settings.
Loading data…
Sustained load · thermal steady state
Batch generation
Maximum simultaneous images at 1024×1024.
Loading data…
Single job · no offload
Benchmark · 05
Video generation.
Clip generation time
Time to render a 5-second 720p clip on an open video model.
Loading data…
Same model file · same prompt
Memory available
Unified/GPU memory available to the video model.
Loading data…
Reference configuration
Energy per clip
Wall-power drawn to render the same 5-second clip.
Loading data…
Wall meter · steady state
Benchmark · 06
Speech recognition.
Word error rate
WER on a mixed English test set with an open ASR model.
Loading data…
Reference test set · greedy decode
Transcription speed
Real-time factor — 1× is real time, higher is faster than real time.
Loading data…
1 hour audio · median of 3
Benchmark · 07
Voice generation.
First audio latency
Time from prompt to first playable audio sample.
Loading data…
Streaming TTS · median of 10
Real-time speed
Real-time factor for continuous voice synthesis.
Loading data…
Sustained synthesis · steady state
Benchmark · 08
Music generation.
Generation time
Time to render a 30-second music clip on an open music model.
Loading data…
Same seed · same prompt
Tracks per hour
Sustained 30-second track throughput per hour.
Loading data…
Continuous load · thermal steady state
Benchmark · 09
Overall AI radar.
Own 1 · Composite AI score
0/10
Composite of the seven benchmark groups above plus cost efficiency. Each axis normalized to a 0–10 scale against the same reference set.
Product
Every angle. Same machine.
Own 1