NVIDIA announces DGX Spark (ships Oct 15)

Made by AIWritten by AI (Claude) from the source below. Check anything important against the source.How it's made

DGX Spark announced (shipping Oct 15) — 128 GB of unified memory

Why it matters

Inference on 200B models and fine-tuning of 70B models on a desktop; local machines become agent runtimes

Source — Primary sources first; secondary reports are labelled “Press”

Officialnvidianews.nvidia.comhttps://nvidianews.nvidia.com/news/nvidia-dgx-spark-arrives-for-worlds-ai-developers

Added in Week 35, 2026 · See that week

On the map — ⑧ Inference Infrastructure line

  1. Previous stopDeepSeek-V3.2-Exp
  2. This recordDGX Spark
  3. Next stopIronwood
See it on the map

Same week

NVFP4 support requested in llama.cpp Discussion #16668OfficialKarpathy appears on Dwarkesh PodcastPressIn the same interview, Karpathy says nanochat was not built by vibe codingPressAnthropic launches Agent Skills (SKILL.md, progressive disclosure)Official

Nearby on the same axis

DeepSeek-V3.2-ExpOfficialGoogle Cloud announces Ironwood (TPU v7)OfficialNVIDIA takes non-exclusive Groq inference license, hires Jonathan Ross and teamOfficialNVIDIA introduces NVFP4Official