Mega Cube

On-device Model Adaptation

This page tracks newly released edge LLMs adapted for MegaCube: each entry ships with a deployment guide and measured benchmark data, so you can pick and roll out on-device models quickly. Click a card to view its adaptation details.

NVIDIA GB10 Grace Blackwell · 1 PFLOPS (FP4) · 128 GB unified memory · 200 G fabric · 1.2 kg