Qualify your hardware
Submit your system profile. Alpha capacity starts at RTX 5090-class hardware or an approved high-memory equivalent.
DACOSAlpha enrollmentApply now
Distributed AI Compute Operating System
Put your GPU to work.
Keep the AI credits.
DACOS connects high-performance personal hardware into a trusted compute network. Share capacity on your terms and earn credits toward private, capable AI.
User-controlled sharing Verified workloads Usage-based credits
The exchange
Your hardware.
A shared advantage.
Choose when to share
You control the availability window. DACOS schedules isolated workloads only while sharing is enabled.
Earn GPU credits
Verified compute time earns platform credits that can be used for AI workloads, including Ultramarine access.
Cost and performance
A different
efficiency curve.
On Paper Instruments' internal PI Knowledge Work benchmark, Ultramarine 1 Flash pairs a competitive max-reasoning score with a materially lower reported task cost.
PI Knowledge WorkMax reasoning
| Model | Benchmark score | Cost per task |
|---|---|---|
| Ultramarine 1 Flash (Max Reasoning) | 59.3 | $0.13 |
| Claude Fable 5 (Max Reasoning) | 57.8 | $8.19 |
| GPT-5.6 Sol (Max Reasoning) | 55.1 | $2.09 |
| Claude Opus 5.5 (Max Reasoning) | ~62.0* | $6.96 |
| GPT-6 Sol (Max Reasoning) | ~62.0* | $1.26 |
* Approximate score. Ultramarine trails these premium models by fewer than three reported points. Figures are supplied by the model developer and have not been independently verified by LinkBox.
Alpha hardware floor
Built for serious compute.
The Alpha begins with an RTX 5090 as the reference minimum. Other architectures are reviewed against usable accelerator memory, sustained throughput, software compatibility, and workload fit—not marketing-tier names alone.
- Stable broadband connection
- Current drivers and supported operating system
- At least 10 available compute hours per week
- Consent to the Alpha compute-sharing terms
PlatformReference systemMemoryReview
NVIDIAGeForce RTX 509032 GB GDDR7Reference minimum
AMDRadeon Pro W790048 GB ECCManual compatibility review
AppleM-series high-memory system128 GB+ unified memoryManual workload review
Multi-GPUProfessional accelerator node48 GB+ available memoryManual topology review
DACOS Alpha
Apply for early access.
Tell us about your primary system. Enrollment is reviewed for hardware eligibility; applying does not install software or begin GPU sharing.
GPU credit policy
Credits accrue from verified, completed compute—not idle time—and have no cash value. Rates and eligible workloads will be disclosed before sharing begins.