从架构上看， MiMo‑V2‑Flash采用了“混合专家模型”（MoE）设计，总参数规模为3090亿，但在实际运行时每次仅激活约150亿参数。 这种方式可以在保证模型能力的同时，大幅降低计算资源消耗。

作者 ｜ 木子今天上午，在 2025 小米人车家全生态合作伙伴大会上，罗福莉首次公开亮相，Title 是 Xiaomi MiMo 大模型负责人。罗福莉还在会上发表演讲，解读了小米的全新大模型 MiMo-V2-Flash ...

在AI领域持续发力的小米公司，近日正式推出了其全新的开源MoE大模型——Xiaomi MiMo-V2-Flash。这个模型的总参数量达到了惊人的3090亿，其中激活参数量为150亿，凭借其卓越的推理效率，迅速引起了行业内的广泛关注。

With Instance, users prompt the AI with their idea, answer a few clarifying questions, and get a complete, working product in minutes. From games and booking systems to custom dashboards and ...

Vienna, Austria, Aug. 08, 2025 (GLOBE NEWSWIRE) -- Mimo, the edtech company that helped over 35 million people learn to code, is launching Instance, an AI-powered tool that lets anyone build fully ...

Chinese AI prodigy Luo Fuli, who recently joined Xiaomi’s MiMo team after a stint at DeepSeek, wrote in a post on X that MiMo-V2-Flash served as “step 2 on our AGI road map”. AGI, the acronym for ...