InfraAug 24, 2026 · 18:448 sources

Xiaomi AI Cube Prototype Runs 120B Model Locally With 1.22TB/s Bandwidth

From huggingnews.com

AI;DR

Xiaomi unveiled the AI Cube, a personal inference hardware prototype that runs large language models locally using three proprietary chips. The system features the O100 accelerator with 1.22TB/s bandwidth and the D100 processor supporting models up to 200B parameters. It incorporates the TSMC-manufactured 3nm Xring O3 SoC and maintains a 150W continuous power release.

Read the sourceSign in to suggest

Sources

+ 6 more
  1. SUPPORT@maxforai“D100 is Xiaomi's smart driving high-performance AI chip, supporting up to 160GB of memory and 200B large model local deployment”x.com
  2. SUPPORT@teortaxestex“6nm, Wafer-on-wafer NPU-DRAM bonding with 1.4µm pitch”x.com
  3. SUPPORT@tonyjzhou“Decoding re-reads active weights every token, so local inference is bandwidth-bound, not FLOPs-bound”x.com
  4. SUPPORT@hesamation“Xiaomi's new prototype looks like a direct competitor to DGX Spark”x.com
  5. SUPPORT@ftr_investors“200K–300K units targeted — reducing its reliance on Qualcomm and MediaTek”x.com
  6. SUPPORT@firstsquawk“Xring D100 3nm smart driving processor”x.com

One ranked edition, every morning

The stories that matter today. Free, cancel anytime.

Telegram @aidr_today