Back to AI intel
趋势

Xiaomi MiMo Claims 1000+ TPS for 1T Model on 8-GPU Server

AI intel briefing

Core summary

One sentence to understand this update

Xiaomi MiMo announced its MiMo-V2.5-Pro UltraSpeed model achieved over 1,000 tokens per second on a 1 trillion parameter MoE model using a single standard 8-GPU server.

Impact & opportunity

What this could mean

This breakthrough from Xiaomi signals new opportunities for builders to deploy large AI models with high performance and cost-effectiveness on standard hardware.