MiMo-V2系列完整指南2026：MiMo-V2-Pro、MiMo-V2-Omni和MiMo-V2-TTS — 小米的智能代理时代AI模型
📄 中文摘要
2026年3月18日，小米推出了三款专门的MiMo-V2模型：MiMo-V2-Pro（推理代理）、MiMo-V2-Omni（全模态基础）和MiMo-V2-TTS（语音合成）。MiMo-V2-Pro在Claw-Eval上得分75.7，全球排名第三，中国排名第二，仅次于Claude Opus 4.6，API成本约为其20%。MiMo-V2-Omni在多模态基准测试中表现出色，包括BigBench Audio（94.0）、MMAU-Pro（69.4）和FutureOmni（66.7）。MiMo-V2-TTS则提供超现实的情感控制、方言合成（四川话、粤语、台语）和准确的唱歌能力。
📄 English Summary
MiMo-V2 Series Complete Guide 2026: MiMo-V2-Pro, MiMo-V2-Omni, and MiMo-V2-TTS — Xiaomi's Agent Era AI Models
On March 18, 2026, Xiaomi launched three specialized MiMo-V2 models: MiMo-V2-Pro (reasoning agent), MiMo-V2-Omni (full-modality base), and MiMo-V2-TTS (speech synthesis). MiMo-V2-Pro scored 75.7 on Claw-Eval, ranking third globally and second in China, just behind Claude Opus 4.6, while costing about 20% of its API price. MiMo-V2-Omni excels in multimodal benchmarks, achieving scores of 94.0 in BigBench Audio, 69.4 in MMAU-Pro, and 66.7 in FutureOmni. MiMo-V2-TTS offers hyper-realistic emotional control, dialect synthesis (Sichuan, Cantonese, Taiwanese), and accurate singing capabilities.