Xiaomi

corporate

china · Founded 2010-04 · IPO 2018 (HKEX: 1810) · Market cap: $109B (HKG: 1810) as of 2026-03

Xiaomi (HKG: 1810) is a global technology company founded in 2010 by Lei Jun, known for smartphones, IoT devices, and the Xiaomi SU7 electric vehicle (360K+ delivered in ~21 months, outselling Tesla Model 3 in China). The Xiao AI voice assistant (100M+ MAU) and HyperOS cross-device operating system underpin Xiaomi's "Human × Car × Home" ecosystem strategy, connecting phones, vehicles, and smart home devices through AI.

The MiMo Team represents Xiaomi's push into frontier AI, led by Luo Fuli, recruited from DeepSeek (reportedly at a salary of "tens of millions of yuan") where he was a key contributor to DeepSeek-V2 and DeepSeek-R1. The team of 63+ researchers has rapidly produced the MiMo model series: MiMo-7B (May 2025, surpassing OpenAI o1-mini on math/code), MiMo-V2-Flash (309B MoE, December 2025), and MiMo-V2-Pro (1T+ parameters, March 2026). MiMo-V2-Pro was revealed to be the mysterious "Hunter Alpha" model that appeared anonymously on OpenRouter, widely misattributed to DeepSeek V4.

Xiaomi also has an emerging robotics program including CyberOne (humanoid) and CyberDog (quadruped), with foundation models driving embodied AI across the "Human × Car × Home" ecosystem.

Website Wikipedia HuggingFace GitHub Artificial Analysis
open-weighton-devicereasoningagenticrobotics

People

Outputs (22)

MiMo-V2-Omni

model

Xiaomi's first fully omni-modal foundation model. Native multimodal architecture that can "see, hear, and act," designed to power Xiaomi's next-genera…

MiMo-V2-Pro

model

Xiaomi's 1-trillion parameter MoE flagship (42B active) designed for agentic workloads with a 1-million-token context window. Before its official reve…

MiMo-V2-TTS

model

High-fidelity text-to-speech model for providing agent voices across Xiaomi's device ecosystem.

Xiaomi-Robotics-0 2

model paper

4.7B parameter Vision-Language-Action (VLA) model for real-time robotic execution. Scales VLA models for consumer GPUs.

PinchBench & ClawEval

dataset

Benchmarks for evaluating multi-step agentic capabilities.

MiMo-V2-Flash 2

model paper

309B MoE model (15B active) optimized for extreme inference speed (150 tokens/sec) and low-cost API access. Introduces a hybrid attention architecture…

↓ 211.4k

MOPD (Multi-Teacher On-Policy Distillation)

library

Library/technique for distilling intelligence from multiple teacher models into smaller student models.

MiMo-Embodied: X-Embodied Foundation Model

paper

First cross-embodied foundation model integrating autonomous driving and embodied AI. SOTA on 17 embodied AI and 12 autonomous driving benchmarks.

↓ 166

Miloco (Xiaomi Local Copilot)

library

Framework enabling LLM-powered smart home automation via local natural language interaction.

HyperOS 3

announcement

First operating system with a system-level AI layer (HyperAI) integrated across smartphones, Xiaomi SU7/SU8 vehicles, and IoT devices.

ACAVCaps

dataset

38,000-hour collection of general audio captions for training holistic audio-reasoning models.

MiMo-VL 2

model paper

Open-source vision-language model that outperformed models ten times its size (like Qwen-72B) in multimodal reasoning. Includes SFT and RL variants.

↓ 1.6k

MiMo-Audio 2

model paper

Specialized model for few-shot audio understanding and environmental sound classification. Pre-trained on 100M+ hours, achieves SOTA on speech intelli…

XRing O1

announcement

3nm custom AI chip designed for on-device reasoning in Xiaomi 15/16/17 series.

MiMo (Reasoning) 2

model paper

Reasoning-focused model achieving superior performance on AIME (math) and HumanEval (coding). From pretraining to posttraining, unlocking reasoning po…

↓ 41.1k

MobileUI Dataset

dataset

Curated sets of mobile application screenshots and UI hierarchies for VLM training.

MobileVLM

model

Vision-language model specialized in understanding mobile UI/UX layouts.

Dasheng 3

paper model

Audio encoder and language model family. Original Dasheng encoder (1.2B params, 272K hours) evolved into DashengLM and MiDashengLM-7B for efficient au…

MiLM-6B

model

Xiaomi's first significant LLM, ranked top in C-Eval and CMMLU at release.

Trustworthy AI White Paper

paper

Outlined Xiaomi's ethical AI principles: Security, Privacy, Fairness.

MACE (Mobile AI Compute Engine)

library

Xiaomi's most famous open-source project. Deep learning inference framework optimized for heterogeneous computing (CPU/GPU/DSP) on mobile devices.

Xiao AI

announcement

Foundational voice assistant that served as the initial testbed for Xiaomi's NLP and speech research.