TL;DR: Xiaomi unveiled an on-premise AI host featuring three new chips (O3, O100, D100) and dual-model support (120B and 3B) with fast/slow inference switching.
Summary: Xiaomi announced a local AI host powered by its newly released O3, O100, and D100 chips. The system supports both a 120B-parameter model and a 3B model, and includes a fast/slow system switching mechanism. It is positioned as an on-premise AI edge computing solution.
Why it matters: This signals a move toward dedicated on-device AI hardware with flexible model scaling for enterprise and edge deployment. AI builders should watch how the fast/slow switching performs in practice and consider testing the dual-model setup for cost-efficient local inference.
Source: x_com