DGX Spark first look: 1.2kg box with 128GB unified memory runs 200B models locally
kimmonismus · x · 2026-09-22
NVIDIA sent DGX Spark to blogger kimmonismus, who shared a first look: 15×15×5.05cm, 1.2kg, powered by the GB10 Grace Blackwell Superchip with 128GB unified CPU/GPU memory and up to 1 petaflop theoretical FP4 sparse compute. NVIDIA claims it can run models up to 200B parameters locally with quantization and fine-tune up to 70B using methods like QLoRA — enabling local fine-tuning, custom image generation, and locally running AI assistants. Hands-on demos to follow.
Related event: NVIDIA DGX Spark Unboxed: 1.2kg Box with 128GB Unified Memory(3 posts)→
More from Infra
- H Company trains computer-use agents on SkyPilot: thousands of sub-second sandboxes — skypilot_org · 2026-09-23
- MiniMax H3 video gen runs locally on M5 Ultra: 768p in ~2m22s with optimizations — bakawolf123 · 2026-09-22
- Cisco: Agentic AI to Drive 9X Enterprise Traffic Growth by 2035 vs 2.5X Without — Beth_Kindig · 2026-09-22
- Cloudflare ships Vary support in Cache Rules to tame HTTP's 'ugliest' header — threepointone · 2026-09-22
- Fluidstack breaks ground on $4B Texas data center campus planned for up to 1.5 GW — MxMnr · 2026-09-22
- Right-size GenAI endpoints: SageMaker concurrency sweeps with Nemotron-3 Nano 30B — AWS ML Blog · 2026-09-22