Unigen Ships M.2 AI Module

What happened

Unigen released the Amaretti E1.S AI module, an M.2-compatible accelerator featuring EdgeCortix's SAKURA-II chip, delivering 60 TOPS (INT8) and up to 32GB of LPDDR4x memory within a 10W power envelope. This module enables local execution of large language models (LLMs) up to 20 billion parameters, supporting GenAI and agentic AI workflows on devices with available M.2 or E1.S slots. The Amaretti E1.S supports major AI frameworks including TensorFlow, PyTorch, ONNX, and Hugging Face, with a stated lead time of 14 weeks.

Why it matters

Deploying frontier models on-device becomes more accessible for platform engineers and solution architects with Unigen's Amaretti module, reducing reliance on cloud infrastructure. The 32GB memory capacity directly addresses the constraint of running 20B parameter LLMs locally, a critical metric for privacy-sensitive or offline applications. This low-power, high-performance solution offers procurement teams a new option for integrating advanced AI capabilities into existing hardware, bypassing typical GPU server lead times with a 14-week delivery.