Xiaomi Showcases Xuanjie O100 Prototype and AI Cube with Self-Developed Chips
Xiaomi has unveiled two hardware prototypes, the Xuanjie O100 mobile terminal and the Xiaomi AI Cube mini PC, powered by its newly announced self-developed Xuanjie chips. The Xuanjie O100 prototype runs the Xiaomi MiMo on-device model, achieving a local inference speed of 295 tokens per second. This development highlights Xiaomi's push toward hardware vertical integration and independent chip design for edge AI. By running large language models locally at high speeds, Xiaomi aims to reduce reliance on cloud-based AI and improve user privacy and latency. The Xuanjie O100 prototype utilizes dual-chip collaborative computing (O3 and O100) and features an active air-cooling system supporting up to 10W of heat dissipation. Meanwhile, the Xiaomi AI Cube integrates the Xuanjie O3, O100, and D100 chips with 80GB of unified memory to run a 120B parameter model alongside a 3B on-device model.
## BACKGROUND
Edge AI refers to running artificial intelligence algorithms locally on hardware devices rather than on centralized cloud servers. Xiaomi's Xuanjie series represents its in-house semiconductor initiative, which includes the O3 flagship SoC, the O100 AI accelerator, and the D100 chip. Xiaomi MiMo is the company's proprietary family of large language models designed for both cloud and on-device applications.