Nov 1, 2024 — by Giorgio Mendoza
57 views
M5Stack has launched the M5Stack LLM Module, an advanced offline large language model inference module designed for terminal devices requiring efficient, cloud-independent AI processing. This product is described as targeting offline applications such as smart homes, voice assistants, and industrial control.
The AX630C SoC appears to include dual-core Arm A53 processors clocked at 1.2GHz, along with a 3.2 TOPs NPU supporting transformer-based models for AI tasks. The module provides 4GB of LPDDR4 memory, with 1GB for user applications and 3GB dedicated to hardware acceleration, alongside 32GB of eMMC storage. Built on a 12nm process, it achieves a runtime power consumption of approximately 1.5W according to the specs.
