标签
XiaomiMiMo 发布 MiMo-V2.5-Pro-FP4-DFlash,这是一款 FP4 量化的 MoE 模型,采用块扩散推测解码,以减少万亿参数推理的内存和带宽。
Xiaomi open-sourced MiMo-V2.5-Pro, a 1.02 trillion parameter MoE model, prompting a cost-benefit analysis of using its API versus self-hosting for autonomous coding tasks.