MiMo-V2.6-Flash, Xiaomi's efficient native omni-modal model in V2.6 series. MoE architecture with 1M-token context window. Enhanced for Agent, tool calling, coding and joint understanding of text/image/video/audio. Outperforms V2.5-Pro, optimized for high-frequency large-scale workloads with great cost performance.