Lenovo previewed the AI mini-host P7 — a 300-gram palm-sized device supporting 7×24 silent desktop operation — that can run a 122-billion-parameter large model locally with power under 30W and noise below 35 dBA, AI Daily reported. It uses a domestic dual-chip architecture and built-in Tianxi AI for local-cloud scheduling, keeping data processed locally for privacy and cost control.
The device reflects rapid progress in edge computing and model compression, making a "desktop private brain" practical for privacy-sensitive scenarios and small-and-medium enterprises without dedicated server rooms. Its small footprint and quiet operation lower the barrier to on-premise AI.
Combined with a wave of low-cost open-source AI hardware — such as Xiaomi's roughly 150-yuan AI desktop pet — China's edge-AI ecosystem is pushing capable models onto inexpensive, personal devices.
The trend points to AI compute migrating from centralized clouds toward distributed, private and affordable endpoints.
Source: AI Daily
