MoDi

This name is used as an internal codename only.

Glenfly HMUA 4.0 Architecture

Unified Training & Inference, Boundless Intelligent Computing

Matrix + Matrix Memory Fabric

Matrix and Matrix Memory Fabric architecture

Matrix Memory Fabric

Matrix Memory Fabric

For intranet AI workstations · private-domain AI training servers

High-performance inference
Large-model training
Private-domain AI

  • DirectX 12 Ultimate

    Higher image quality and faster performance

  • Matrix Core

    Dedicated matrix acceleration for a dual boost in AI computing power and energy efficiency

  • Matrix Memory Fabric

    Direct compute-memory interconnect for ultimate data sharing and flow

  • Low-power SLM engine

    dedicated collaborative optimization for multiple lightweight small models

  • Agent scheduling engine 2.0

    Dynamically orchestrates agent data interconnection to improve complex-task execution efficiency

  • Glenfly HUCA 2.0 unified heterogeneous computing

    Unified training & inference with full-stack efficient adaptation and deployment