Built on Qwen2.5-Omni-7B, it is embedded in Xiaomi's electric vehicles and smart home devices, released under Apache 2.0 license.
MiMo-Audio is a separate audio language model whose encoder was integrated into MiMo-V2.5 for omnimodal capabilities.
OmniVoice: Voice Cloning in 646 Languages
OmniVoice, open-sourced in May 2026, is a zero-shot voice cloning TTS model supporting 646 languages.
It can clone a voice from a few seconds of audio and generate natural speech across languages while preserving voice characteristics.
Its single-transformer architecture maps text directly to acoustic tokens, completing training on 100,000 hours of audio in a day and running inference at up to 40x real-time speed.
Xiaomi claims it is the first voice cloning TTS model covering hundreds of languages, released under Apache-2.0.
Alongside V2.5, Xiaomi released MiMo-V2.5-TTS and an ASR system for bilingual recognition, enabling end-to-end voice-driven products.
Consumer AI: Xiao AI and HyperAI
Xiao AI (小爱) is Xiaomi's voice assistant, upgraded to "Super Xiao AI" with HyperOS 2, featuring better context memory, smart home control, and text-to-image generation.
>>> Moto G Max Launched with 200MP Camera, 5000-nit Display, and Rugged Build
HyperAI, introduced globally at MWC 2025, includes real-time translation, AI writing, speech recognition, and AI photo editing.
It also integrates Google Gemini for global devices and has expanded to mid-range phones like Redmi Note 14 Pro+ 5G and Poco series.
miclaw: Autonomous AI Agent
Announced in March 2026 and in closed beta, miclaw is an autonomous AI agent built on MiMo.
It interprets user intent and performs multi-step tasks across apps, interfaces, and system tools without supervision. It uses an "inference-execution loop" and has contextual memory.
User interactions are not used for training; sensitive data is processed locally via edge-cloud privacy computing.
The closed beta supports Xiaomi 17 series, and HyperOS 4 will fully integrate miclaw at the system level.
It has also been tested as a smartwatch assistant.
Investment and Impact
In March 2026, CEO Lei Jun announced at least $8.7 billion in AI investment over three years.
By April 2026, Xiaomi's models captured about 21% of all traffic on OpenRouter.
The company aims for a "grand convergence" in 2026, combining its own chip, OS, and AI model in a single device.
Xiaomi now has a full AI stack: reasoning models, vision-language models, audio models, voice cloning, TTS/ASR pipeline, an AI agent, and consumer features on millions of devices.
Most models are open-source, building developer momentum.
>>> Anker launches 160W smart charger with Huawei 66W and Xiaomi 120W fast charging
The key test ahead is whether miclaw and HyperOS 4 make AI truly useful in daily life, potentially positioning Xiaomi as a genuine AI platform.
