MiMo-V2-Omni
Xiaomi · 2026-03-18
MiMo-V2-Omni was Xiaomi's closed, API-hosted omni-modal foundation model, released March 18, 2026 for unified text, image, video, and audio understanding; Xiaomi routed the API identifier to MiMo-V2.5 from June 1 and deprecated it on June 30, 2026. It supported up to a 256K context window and native structured tool calling, function execution, and UI grounding, pairing multimodal perception with agentic planning and action.