Sinun AI
fi

Hardware

DGX Spark — a local AI supercomputer on your desk

Nvidia's GB10 Grace Blackwell-based DGX Spark shipped in October 2025. 1 PFLOP FP4, 128 GB unified memory, $3,999 — runs models up to 200B parameters locally.

1 PFLOP FP4128 GBUp to 200B

DGX Spark is the former Project DIGITS — Nvidia's answer to the personal AI workstation market. It's the most powerful local AI machine you can order by mail, ideal for larger models, local fine-tuning and research.

Who is DGX Spark for?

Researchers, dev teams and regulated sectors doing local fine-tuning or heavy inference as daily work. Companies where a $4,000 device is cheap compared to a monthly cloud bill.

DGX Spark vs Framework Desktop

DGX Spark wins raw performance: image gen ~120 TFLOPS (Strix Halo ~46), prompt processing 1,723 tokens/s (Strix Halo 340). Framework doubles price-performance. We recommend DGX only when you really need it.

Frequently asked

Can I get one in Finland?
Yes, Nvidia partners ship to Finland. We're happy to help with procurement, installation and onboarding.

Updated 2026-04-21

Related

Hardware

Mac Studio — the quiet default for local AI

Apple's Mac Studio (M4 Max / M4 Ultra) is one of 2026's best local-AI workstations. Up to 512 GB of unified memory, silent, power-efficient.

Hardware

Framework Desktop — Strix Halo, 128 GB, ~$2,400

The Framework Desktop with AMD Ryzen AI Halo delivers 128 GB of unified memory at about $2,400. The best price-to-performance workstation for local AI in 2026.

Hardware

MacBook — enough for most, travels with you

A MacBook Pro (M4 Pro/Max) runs most mid-size open LLMs smoothly in 2026. When is a laptop enough, and when do you need a workstation?

All resources

Want your own local AI assistant?

Tell us about your work and hardware — we'll map the right model, the right hardware tier and the right sync configuration.

Get in Touch