comscore

Xiaomi Launches MiDashengLM-7B AI Model with Voice Control for Smart Homes and Cars, Offers Open Source Access for Developers

Xiaomi has launched MiDashengLM-7B, an advanced AI voice model now powering smart homes and cars. With open-source access, multi-sound recognition, and low-latency performance, it’s built for real-world use and developer innovation.

Published By: Deepti Ratnam | Published: Aug 05, 2025, 11:15 AM (IST)

  • whatsapp
  • twitter
  • facebook
  • whatsapp
  • twitter
  • facebook

Xiaomi has leaped into Artificial Intelligence by debuting its voice-based model, dubbed MiDashengLM-7B. The model is live and active in Xiaomi’s smart home products, including smart TVs and vehicles. Xiaomi’s new AI model is available in China right now, and the company has open-sourced it for developers and businesses. The MiDashengLM-7B voice-based AI model is available under the Apache 2.0 license in China. It will allow developers and businesses to integrate the model into their own applications. It is suitable for both research and commercial use.

What is Xiaomi’s new MiDashengLM-7B AI Model?

The new Xiaomi MiDashengLM-7B AI Model doesn’t work like other voice assistants, and hence, it is trained by using the public available datasets. What makes this model different from other voice assistants is that it is not just capable of interpreting human speech, but also capable enough to translate and transcribe environmental and ambient sound.

The company combined Xiaomi’s Dasheng audio encoder with Alibaba’s Qwen2.5-Omni-7B decoder to create this model. This collaboration and technological advancement allow this model to respond to sounds like clap, snaps, and even detect noises in a room. The model is suitable for security alerts and gesture-based controls.

Real-Time Capabilities in Homes and Cars

Xiaomi has rolled out its MiDashengLM-7B AI Model with more than 30 AI-powered features. These features will be embedded across company’s ecosystem. Additionally, the model enables constant audio surveillance, which allows it to work like a safety guard. Other than this, it can also send alerts if something unusual is detected.

Talking about vehicles, the model delivers voice-activated commands along with providing language pronunciation feedback. One unique feature includes underwater voice wake-up, where sound replaces touch input that is perfect for wet environments.

As per company, the MiDashengLM-7B AI Model responds faster and handles 2X more tasks simultaneously. It doesn’t require extra money or cloud servers, making it runs smoothly without worry about being connected to the internet.